The first author acknowledges support from CNPq Projeto Universal 408687/2023-1 “Geometria das Equações Diferenciais Algébricas” and from CNPq-Brazil PQ-306011/2023-9. The second author is supported by the Spanish grant PID2023-149508NB-I00, funded by MICIU/AEI/10.13039/501100011033 and by FEDER, UE. The third author acknowledges support from CAPES, Brazil. The first and fourth authors were also supported by the Vicerrectorado de Investigación (VRI) of the Pontificia Universidad Católica del Perú (PUCP) through grant DFI-2025-PI1275.
Introduction
Let O n = C { x 1 , … , x n } \mathcal{O}_n = \mathbb{C}\{x_1,\ldots,x_n\} O n = C { x 1 , … , x n } denote the ring of germs of holomorphic functions at the origin of C n \mathbb{C}^n C n , with maximal ideal m \mathfrak{m} m . Let f ∈ m f \in\mathfrak{m} f ∈ m define an isolated hypersurface singularity ( V , 0 ) = ( f − 1 ( 0 ) , 0 ) (V,0) = (f^{-1}(0),0) ( V , 0 ) = ( f − 1 ( 0 ) , 0 ) , and denote by J ( f ) = ( f x 1 , … , f x n ) J(f) = (f_{x_1},\ldots,f_{x_n}) J ( f ) = ( f x 1 , … , f x n ) its Jacobian ideal. The Milnor and Tjurina numbers of f f f are defined by μ ( f ) = dim C O n / J ( f ) \mu(f) = \dim_{\mathbb{C}} \mathcal{O}_n/J(f) μ ( f ) = dim C O n / J ( f ) and τ ( f ) = dim C O n / ( ( f ) + J ( f ) ) \tau(f) = \dim_{\mathbb{C}} \mathcal{O}_n/((f) + J(f)) τ ( f ) = dim C O n / (( f ) + J ( f )) , respectively. The Milnor number is the rank of the middle homology of the Milnor fiber of f f f [19 ] , whereas the Tjurina number is the dimension of the base of a semi-universal deformation of ( V , 0 ) (V,0) ( V , 0 ) [26 ] . One always has τ ( f ) ≤ μ ( f ) \tau(f) \leq\mu(f) τ ( f ) ≤ μ ( f ) , and Saito’s theorem [22 ] states that equality holds if and only if f f f is quasihomogeneous up to a change of coordinates. Moreover, the Mather–Yau theorem [18 ] asserts that the algebra O n / ( ( f ) + J ( f ) ) \mathcal{O}_n/((f) + J(f)) O n / (( f ) + J ( f )) determines the analytic type of ( V , 0 ) (V,0) ( V , 0 ) .
The algebras O n / m k J ( f ) \mathcal{O}_n/\mathfrak{m}^kJ(f) O n / m k J ( f ) and O n / ( ( f ) + m k J ( f ) ) \mathcal{O}_n/((f) + \mathfrak{m}^kJ(f)) O n / (( f ) + m k J ( f )) were first considered by Greuel–Pham [10 ] . The corresponding k k k -th Milnor and Tjurina numbers were subsequently studied by Hussain–Liu–Yau–Zuo [13 ] and Hu–Yau–Zuo [12 ] . For each integer k ≥ 0 k \geq0 k ≥ 0 , one considers M k ( f ) = O n / m k J ( f ) M_k(f) = \mathcal{O}_n/\mathfrak{m}^kJ(f) M k ( f ) = O n / m k J ( f ) and T k ( f ) = O n / ( ( f ) + m k J ( f ) ) T_k(f) = \mathcal{O}_n/((f) + \mathfrak{m}^kJ(f)) T k ( f ) = O n / (( f ) + m k J ( f )) . Their dimensions μ k ( f ) = dim C M k ( f ) \mu^k(f) = \dim_{\mathbb{C}} M_k(f) μ k ( f ) = dim C M k ( f ) and τ k ( f ) = dim C T k ( f ) \tau^k(f) = \dim_{\mathbb{C}} T_k(f) τ k ( f ) = dim C T k ( f ) are called the k k k -th Milnor number and the k k k -th Tjurina number of f f f , respectively. For k = 0 k = 0 k = 0 , these are the classical Milnor and Tjurina numbers, and τ k ( f ) ≤ μ k ( f ) \tau^k(f) \leq\mu^k(f) τ k ( f ) ≤ μ k ( f ) for every k ≥ 0 k \geq0 k ≥ 0 .
The relation between these numbers differs from the classical case. If f f f is quasihomogeneous, the Euler identity gives f ∈ m J ( f ) f \in\mathfrak{m}J(f) f ∈ m J ( f ) , so τ 1 ( f ) = μ 1 ( f ) \tau^1(f) = \mu^1(f) τ 1 ( f ) = μ 1 ( f ) . For k ≥ 2 k \geq2 k ≥ 2 , however, f f f need not belong to m k J ( f ) \mathfrak{m}^kJ(f) m k J ( f ) , and the two numbers may differ even for a quasihomogeneous singularity. Questions concerning the quotient μ k ( f ) / τ k ( f ) \mu^k(f)/\tau^k(f) μ k ( f ) / τ k ( f ) and the relation between the k k k -th Milnor number of a semiquasihomogeneous germ and that of its quasihomogeneous principal part have led to several conjectures [14 , 23 ] .
In this paper, we address these questions and investigate the corresponding invariants for pairs of plane holomorphic germs. Among our main contributions, we provide counterexamples to two conjectures concerning isolated hypersurface singularities, while establishing the validity of one of them for plane curve singularities. We also derive explicit formulas for the k k k -th Milnor and Tjurina numbers of pairs and explore their relationship with the bifurcation fibers of a pencil of plane curves. Our approach relies on the extension of these invariants to germs of holomorphic foliations in dimension two, developed by Ribeiro [20 ] in her doctoral thesis and presented in [21 ] . In particular, by applying these results to the foliation defined by ω f / g = g d f − f d g \omega_{f/g} = g\,df - f\,dg ω f / g = g df − f d g , we obtain the desired formulas for pairs. A more detailed study of the Milnor and Tjurina numbers of foliations on ( C 2 , 0 ) (\mathbb{C}^{2},0) ( C 2 , 0 ) can be found in [7 ] .
The paper is organized as follows. § 2 recalls the necessary definitions and preliminary results.
In § 3 , we investigate a conjecture of Hussain–Yau–Zuo [14 Conjecture 1.4] , which predicts that μ k ( f ) / τ k ( f ) < 3 / 2 \mu^{k}(f)/\tau^{k}(f)<3/2 μ k ( f ) / τ k ( f ) < 3/2 for every k ≥ 0 k\geq0 k ≥ 0 and every isolated hypersurface singularity V = { f = 0 } ⊂ ( C 3 , 0 ) V=\{f=0\}\subset(\mathbb{C}^{3},0) V = { f = 0 } ⊂ ( C 3 , 0 ) . The case k = 0 k=0 k = 0 is precisely a conjecture of Almirón [1 Conjecture 4.10] , which was recently proved by Ma–Zuo [17 Corollary 4.8] . We show that the bound does not hold in general for k ≥ 1 k\geq1 k ≥ 1 . Already for f = x 2 + y 2 + z 2 f=x^{2}+y^{2}+z^{2} f = x 2 + y 2 + z 2 one has μ k ( f ) = ( k + 3 3 ) \mu^{k}(f)=\binom{k+3}{3} μ k ( f ) = ( 3 k + 3 ) and τ k ( f ) = ( k + 1 ) 2 \tau^{k}(f)=(k+1)^{2} τ k ( f ) = ( k + 1 ) 2 , so μ k ( f ) / τ k ( f ) > 3 / 2 \mu^{k}(f)/\tau^{k}(f)>3/2 μ k ( f ) / τ k ( f ) > 3/2 exactly for k ≥ 5 k\geq5 k ≥ 5 ; for k = 5 k=5 k = 5 the quotient is 14 / 9 14/9 14/9 . More generally, for every isolated hypersurface singularity in n ≥ 2 n\geq2 n ≥ 2 variables, μ k / τ k \mu^{k}/\tau^{k} μ k / τ k tends to infinity as k → ∞ k\to\infty k → ∞ ; see Theorem 3.5 . In particular, the conjecture fails for the families considered by Hussain–Yau–Zuo, who checked only k ≤ 2 k\leq2 k ≤ 2 . For instance, for f = x 7 + y 7 + z 7 + x 3 y 3 z 3 f=x^{7}+y^{7}+z^{7}+x^{3}y^{3}z^{3} f = x 7 + y 7 + z 7 + x 3 y 3 z 3 , computations performed with Singular [5 ] yield μ 16 ( f ) = 2024 \mu^{16}(f)=2024 μ 16 ( f ) = 2024 and τ 16 ( f ) = 1344 \tau^{16}(f)=1344 τ 16 ( f ) = 1344 , hence μ 16 ( f ) / τ 16 ( f ) = 253 / 168 > 3 / 2 \mu^{16}(f)/\tau^{16}(f)=253/168>3/2 μ 16 ( f ) / τ 16 ( f ) = 253/168 > 3/2 . The analogous conjecture for plane curves [14 Conjecture 1.3] also fails.
§ 4 is devoted to a conjecture of Shi–Yau–Zuo [23 Conjecture 4.65] on the behaviour of the k k k -th Milnor numbers under semiquasihomogeneous perturbations. More precisely, the conjecture asserts that, if f = f w + f ′ f=f_{w}+f' f = f w + f ′ is semiquasihomogeneous with quasihomogeneous principal part f w f_{w} f w , then μ k ( f ) = μ k ( f w ) \mu^{k}(f)=\mu^{k}(f_{w}) μ k ( f ) = μ k ( f w ) for every k ≥ 0 k\geq0 k ≥ 0 . Shi–Yau–Zuo established this equality under the additional assumption k ≤ min i { ν ( f x i ) } k\leq\min_{i}\{\nu(f_{x_{i}})\} k ≤ min i { ν ( f x i )} [23 Theorem C] . We prove that the conjecture holds for all k ≥ 0 k\geq0 k ≥ 0 in the case of plane curve singularities, that is, when f ∈ C [ [ x , y ] ] f\in\mathbb{C[[x,y]]} f ∈ C [[ x , y ]] (and hence also when f ∈ C { x , y } f\in\mathbb{C\{x,y\}} f ∈ C { x , y } ); see Theorem 4.2 . Our proof relies on a formula expressing μ k ( f ) \mu^{k}(f) μ k ( f ) in terms of the classical Milnor number and the multiplicity of f f f [21 Corollary 4.1] , together with the equality μ ( f ) = μ ( f w ) \mu(f)=\mu(f_{w}) μ ( f ) = μ ( f w ) , which holds for convergent series by [2 ] and for formal series by [4 Proposition 2] . In contrast, we show that the conjecture fails in dimension three. To this end, we consider f = x 3 + y 7 + z 7 + x y 2 z 3 f=x^{3}+y^{7}+z^{7}+xy^{2}z^{3} f = x 3 + y 7 + z 7 + x y 2 z 3 and f w = x 3 + y 7 + z 7 f_{w}=x^{3}+y^{7}+z^{7} f w = x 3 + y 7 + z 7 , where f w f_{w} f w is quasihomogeneous with weights ( 7 , 3 , 3 ) (7,3,3) ( 7 , 3 , 3 ) . In this case, μ 6 ( f ) = 199 \mu^{6}(f)=199 μ 6 ( f ) = 199 whereas μ 6 ( f w ) = 200 \mu^{6}(f_{w})=200 μ 6 ( f w ) = 200 , providing an explicit counterexample. In a recent preprint, Liu–Yau–Zhang [16 Main Theorem A] proved the equal-weight case of the conjecture for convergent germs in every dimension n ≥ 2 n\geq2 n ≥ 2 , namely for germs whose homogeneous principal part defines a smooth projective hypersurface, equivalently has an isolated critical point at the origin. Their result complements Theorem 4.2 , which allows arbitrary positive weights in dimension two. Our three-variable counterexample has weights ( 7 , 3 , 3 ) (7,3,3) ( 7 , 3 , 3 ) and therefore lies outside their setting. Their study of higher Milnor sequences also uses the plane curve formula of Ribeiro and Fernández-Pérez [21 Corollary 4.1] .
In § 5 , following [8 , 21 ] , we introduce the k k k -th Milnor and Tjurina numbers of a pair of plane holomorphic germs. Let f , g ∈ m ⊂ O 2 f,g\in\mathfrak{m}\subset\mathcal{O}_{2} f , g ∈ m ⊂ O 2 be reduced and relatively prime, and assume that the Milnor number of the pair, defined in [25 ] , satisfies
μ ( f , g ) = dim C O 2 ( g f x − f g x , g f y − f g y ) < ∞ . \mu(f,g)=\dim_{\mathbb{C}}\frac{\mathcal{O}_2}{(gf_x-fg_x,gf_y-fg_y)}<\infty. μ ( f , g ) = dim C ( g f x − f g x , g f y − f g y ) O 2 < ∞.
For each integer k ≥ 0 k\geq0 k ≥ 0 , we define
μ k ( f , g ) = dim C O 2 m k ( g f x − f g x , g f y − f g y ) . \mu^k(f,g)=\dim_{\mathbb{C}}\frac{\mathcal{O}_2}{\mathfrak{m}^k(gf_x-fg_x,gf_y-fg_y)}. μ k ( f , g ) = dim C m k ( g f x − f g x , g f y − f g y ) O 2 .
For a member C [ α : β ] : α f + β g = 0 C_{[\alpha:\beta]}:\alpha f+\beta g=0 C [ α : β ] : α f + β g = 0 of the pencil generated by f f f and g g g , we also define
τ [ α : β ] k ( f , g ) = dim C O 2 m k ( g f x − f g x , g f y − f g y ) + ( α f + β g ) . \tau^k_{[\alpha:\beta]}(f,g)=\dim_{\mathbb{C}}\frac{\mathcal{O}_2}{\mathfrak{m}^k(gf_x-fg_x,gf_y-fg_y)+(\alpha f+\beta g)}. τ [ α : β ] k ( f , g ) = dim C m k ( g f x − f g x , g f y − f g y ) + ( α f + β g ) O 2 .
Theorem 5.1 expresses μ k ( f , g ) \mu^k(f,g) μ k ( f , g ) in terms of μ ( f , g ) \mu(f,g) μ ( f , g ) , k k k , and the algebraic multiplicity of the 1 1 1 -form ω f / g = g d f − f d g \omega_{f/g}=g\,df-f\,dg ω f / g = g df − f d g . For every member C [ α : β ] C_{[\alpha:\beta]} C [ α : β ] of the pencil, which is reduced by [6 Lemma 4.1] , Proposition 5.2 gives τ [ α : β ] k ( f , g ) = τ k ( C [ α : β ] ) + I 0 ( f , g ) \tau^k_{[\alpha:\beta]}(f,g)=\tau^k(C_{[\alpha:\beta]})+I_0(f,g) τ [ α : β ] k ( f , g ) = τ k ( C [ α : β ] ) + I 0 ( f , g ) , where I 0 ( f , g ) I_0(f,g) I 0 ( f , g ) denotes the intersection multiplicity of f = 0 f=0 f = 0 and g = 0 g=0 g = 0 at the origin. Thus the difference between these two Tjurina numbers is independent of both k k k and the chosen member of the pencil. Combining this with the invariant η ( f , g ) \eta(f,g) η ( f , g ) introduced by Suwa [24 ] , Theorem 5.4 gives, for every member C [ α : β ] C_{[\alpha:\beta]} C [ α : β ] of the pencil and every k ≥ 0 k\geq0 k ≥ 0 ,
μ k ( f , g ) − τ [ α : β ] k ( f , g ) = η ( f , g ) + k ( k + 1 ) − τ k ( C [ α : β ] ) − ε k ( m 0 ) , \mu^k(f,g)-\tau^k_{[\alpha:\beta]}(f,g)=\eta(f,g)+k(k+1)-\tau^k(C_{[\alpha:\beta]})-\varepsilon_k(m_0), μ k ( f , g ) − τ [ α : β ] k ( f , g ) = η ( f , g ) + k ( k + 1 ) − τ k ( C [ α : β ] ) − ε k ( m 0 ) ,
where m 0 m_0 m 0 is the algebraic multiplicity of ω f / g \omega_{f/g} ω f / g and ε k ( m 0 ) \varepsilon_k(m_0) ε k ( m 0 ) is the correction term of Theorem 5.1 .
Finally, in Section 6, we compare μ k ( f , g ) \mu^k(f,g) μ k ( f , g ) with μ k ( f g ) \mu^k(fg) μ k ( f g ) . Starting from the bifurcation formula [6 Theorem 4.6] , we express their difference as the sum of the Milnor number jumps at the bifurcation fibers distinct from f = 0 f=0 f = 0 and g = 0 g=0 g = 0 , together with an explicit correction depending on k k k and the multiplicities of f f f , g g g , and ω f / g \omega_{f/g} ω f / g , namely ε k ( M − 1 ) − ε k ( m 0 ) \varepsilon_k(M-1)-\varepsilon_k(m_0) ε k ( M − 1 ) − ε k ( m 0 ) with M = ν ( f ) + ν ( g ) M=\nu(f)+\nu(g) M = ν ( f ) + ν ( g ) and m 0 = ν ( ω f / g ) m_0=\nu(\omega_{f/g}) m 0 = ν ( ω f / g ) ; see Theorem 6.2 . Moreover, m 0 = M − 1 + δ m_0=M-1+\delta m 0 = M − 1 + δ , where δ \delta δ is the sum, over the bifurcation fibers distinct from f = 0 f=0 f = 0 and g = 0 g=0 g = 0 , of the excess of their multiplicity over the generic one. It vanishes unless ν ( f ) = ν ( g ) \nu(f)=\nu(g) ν ( f ) = ν ( g ) and the initial forms of f f f and g g g are proportional, and then only one fiber contributes; see Proposition 6.1 . When δ = 0 \delta=0 δ = 0 , that is, when ν ( ω f / g ) = ν ( f ) + ν ( g ) − 1 \nu(\omega_{f/g})=\nu(f)+\nu(g)-1 ν ( ω f / g ) = ν ( f ) + ν ( g ) − 1 , the formula reduces to
μ k ( f , g ) − μ k ( f g ) = ∑ [ α : β ] ∈ B ∗ ( f , g ) ( μ ( C [ α : β ] ) − μ g e n ) for every k ≥ 0. \mu^k(f,g)-\mu^k(fg)=\sum_{[\alpha:\beta]\in\mathcal{B}^*(f,g)}\left(\mu(C_{[\alpha:\beta]})-\mu_{\mathrm{gen}}\right)\quad\text{for every }k\geq0. μ k ( f , g ) − μ k ( f g ) = [ α : β ] ∈ B ∗ ( f , g ) ∑ ( μ ( C [ α : β ] ) − μ gen ) for every k ≥ 0.
Here μ g e n \mu_{\mathrm{gen}} μ gen is the Milnor number of a generic member of the pencil, and B ∗ ( f , g ) \mathcal{B}^*(f,g) B ∗ ( f , g ) is the set of bifurcation parameters different from [ 1 : 0 ] [1:0] [ 1 : 0 ] and [ 0 : 1 ] [0:1] [ 0 : 1 ] . In this case, this difference is independent of k k k . In general, the meromorphic germ ( f : g ) (f:g) ( f : g ) is semitame if and only if μ k ( f , g ) = μ k ( f g ) \mu^k(f,g)=\mu^k(fg) μ k ( f , g ) = μ k ( f g ) for some, equivalently every integer k ≥ 0 k\geq0 k ≥ 0 ; see Corollary 6.5 .
Basic definitions and preliminary results
Let ( V , 0 ) ⊂ ( C n , 0 ) (V,0) \subset(\mathbb{C}^{n},0) ( V , 0 ) ⊂ ( C n , 0 ) be an isolated hypersurface singularity defined by a germ f ∈ O n = C { x 1 , … , x n } f \in\mathcal{O}_{n} = \mathbb{C}\{x_{1},\ldots,x_{n}\} f ∈ O n = C { x 1 , … , x n } , and let m \mathfrak{m} m denote the maximal ideal of O n \mathcal{O}_{n} O n . Denote by ν ( f ) \nu(f) ν ( f ) the order of f ∈ O n f \in\mathcal{O}_{n} f ∈ O n . Following Hussain–Liu–Yau–Zuo [13 ] , for each integer k ≥ 0 k \ge0 k ≥ 0 we define the k k k -th Milnor number and the k k k -th Tjurina number of f f f by
μ k ( f ) = dim C O n m k J ( f ) , τ k ( f ) = dim C O n ( f ) + m k J ( f ) . \mu^{k}(f)=\dim_{\mathbb{C}}\frac{\mathcal{O}_{n}}{\mathfrak{m}^{k}J(f)},\qquad\tau^{k}(f)=\dim_{\mathbb{C}}\frac{\mathcal{O}_{n}}{(f)+\mathfrak{m}^{k}J(f)}. μ k ( f ) = dim C m k J ( f ) O n , τ k ( f ) = dim C ( f ) + m k J ( f ) O n .
where J ( f ) = ( f x 1 , … , f x n ) J(f)=(f_{x_{1}},\ldots,f_{x_{n}}) J ( f ) = ( f x 1 , … , f x n ) is the Jacobian ideal of f f f .
In Ribeiro’s PhD thesis [20 ] (see also [21 ] ), these notions were extended to germs of holomorphic foliations in dimension two. More precisely, let ω = P ( x , y ) d x + Q ( x , y ) d y \omega=P(x,y)\,dx+Q(x,y)\,dy ω = P ( x , y ) d x + Q ( x , y ) d y be a germ of holomorphic 1-form defining a holomorphic foliation F \mathcal{F} F on ( C 2 , 0 ) (\mathbb{C}^{2},0) ( C 2 , 0 ) with an isolated singularity at the origin, that is, P , Q ∈ m P,Q\in\mathfrak{m} P , Q ∈ m have no common factor. By definition, the algebraic multiplicity of F \mathcal{F} F is ν ( F ) : = min { ν ( P ) , ν ( Q ) } \nu(\mathcal{F}):=\min\{\nu(P),\nu(Q)\} ν ( F ) := min { ν ( P ) , ν ( Q )} , and we also write ν ( ω ) = ν ( F ) \nu(\omega)=\nu(\mathcal{F}) ν ( ω ) = ν ( F ) .
For f , g ∈ O 2 f,g\in\mathcal{O}_{2} f , g ∈ O 2 , the intersection multiplicity of f f f and g g g at the origin is I 0 ( f , g ) : = dim C ( O 2 / ( f , g ) ) I_{0}(f,g):=\dim_{\mathbb{C}}(\mathcal{O}_{2}/(f,g)) I 0 ( f , g ) := dim C ( O 2 / ( f , g )) , where ( f , g ) (f,g) ( f , g ) denotes the ideal generated by f f f and g g g . By definition, the Milnor number of f ∈ O 2 f\in\mathcal{O}_{2} f ∈ O 2 is μ ( f ) = I 0 ( f x , f y ) \mu(f)=I_{0}(f_{x},f_{y}) μ ( f ) = I 0 ( f x , f y ) , which coincides with dim C O 2 / J ( f ) \dim_{\mathbb{C}}\mathcal{O}_{2}/J(f) dim C O 2 / J ( f ) , and the Milnor number of the foliation F \mathcal{F} F is μ ( F ) = I 0 ( P , Q ) \mu(\mathcal{F})=I_{0}(P,Q) μ ( F ) = I 0 ( P , Q ) .
Let C = { f = 0 } C=\{f=0\} C = { f = 0 } be a germ of a complex reduced curve, where f ∈ m f\in\mathfrak{m} f ∈ m . The curve C C C is said to be invariant by F \mathcal{F} F or F \mathcal{F} F -invariant if ( ω ∧ d f ) / f ∈ Ω 2 ( C 2 , 0 ) (\omega\wedge df)/f\in\Omega^{2}(\mathbb{C}^{2},0) ( ω ∧ df ) / f ∈ Ω 2 ( C 2 , 0 ) , where Ω 2 ( C 2 , 0 ) \Omega^{2}(\mathbb{C}^{2},0) Ω 2 ( C 2 , 0 ) is the module of germs of holomorphic 2-forms on ( C 2 , 0 ) (\mathbb{C}^{2},0) ( C 2 , 0 ) .
The k k k -th Milnor number of F \mathcal{F} F and the k k k -th Tjurina number of F \mathcal{F} F along C C C are then defined by
μ k ( F ) = dim C O 2 m k ( P , Q ) , τ k ( F , C ) = dim C O 2 m k ( P , Q ) + ( f ) . \mu^{k}(\mathcal{F})=\dim_{\mathbb{C}}\frac{\mathcal{O}_{2}}{\mathfrak{m}^{k}(P,Q)},\qquad\tau^{k}(\mathcal{F},C)=\dim_{\mathbb{C}}\frac{\mathcal{O}_{2}}{\mathfrak{m}^{k}(P,Q)+(f)}. μ k ( F ) = dim C m k ( P , Q ) O 2 , τ k ( F , C ) = dim C m k ( P , Q ) + ( f ) O 2 .
respectively. For ω = d f \omega=df ω = df , we have ( P , Q ) = J ( f ) (P,Q)=J(f) ( P , Q ) = J ( f ) , so μ k ( F ) = μ k ( f ) \mu^{k}(\mathcal{F})=\mu^{k}(f) μ k ( F ) = μ k ( f ) and τ k ( F , C ) = τ k ( f ) \tau^{k}(\mathcal{F},C)=\tau^{k}(f) τ k ( F , C ) = τ k ( f ) .
The following corollary relates the k k k -th Milnor number of a plane curve to its Milnor number.
Corollary 2.1 ([21 Corollary 4.1] ). Let f ∈ m ⊂ O 2 f\in\mathfrak{m}\subset\mathcal{O}_{2} f ∈ m ⊂ O 2 define an isolated singularity at the origin 0 ∈ C 2 0\in\mathbb{C}^{2} 0 ∈ C 2 , that is, μ ( f ) < ∞ \mu(f)<\infty μ ( f ) < ∞ , and let m = ν ( f ) m=\nu(f) m = ν ( f ) be the order of f f f . Then
μ k ( f ) = { μ ( f ) + k ( k + 1 ) , if 0 ≤ k < m , μ ( f ) + k ( k + 1 ) − ( k − m + 2 ) ( k − m + 1 ) 2 , if k ≥ m . \mu^{k}(f)=
\begin{cases}
\mu(f)+k(k+1), & \text{if } 0\le k<m,\\
\mu(f)+k(k+1)-\dfrac{(k-m+2)(k-m+1)}{2}, & \text{if } k\ge m.
\end{cases} μ k ( f ) = ⎩ ⎨ ⎧ μ ( f ) + k ( k + 1 ) , μ ( f ) + k ( k + 1 ) − 2 ( k − m + 2 ) ( k − m + 1 ) , if 0 ≤ k < m , if k ≥ m .
Applying Corollary 2.1 to the product f g fg f g , we obtain
Corollary 2.2. Let f , g ∈ m ⊂ O 2 f,g \in\mathfrak{m} \subset\mathcal{O}_2 f , g ∈ m ⊂ O 2 be reduced and coprime and let m = ν ( f ) m=\nu(f) m = ν ( f ) and ℓ = ν ( g ) \ell=\nu(g) ℓ = ν ( g ) be the orders of f f f and g g g . Then
μ k ( f g ) = { μ ( f g ) + k ( k + 1 ) , if 0 ≤ k < m + ℓ , μ ( f g ) + k ( k + 1 ) − ( k − ( m + ℓ ) + 2 ) ( k − ( m + ℓ ) + 1 ) 2 , if k ≥ m + ℓ . \mu^k(fg)=
\begin{cases}
\mu(fg)+k(k+1), & \text{if } 0\leq k<m+\ell,\\
\mu(fg)+k(k+1)-\dfrac{(k-(m+\ell)+2)(k-(m+\ell)+1)}{2}, & \text{if } k\geq m+\ell.
\end{cases} μ k ( f g ) = ⎩ ⎨ ⎧ μ ( f g ) + k ( k + 1 ) , μ ( f g ) + k ( k + 1 ) − 2 ( k − ( m + ℓ ) + 2 ) ( k − ( m + ℓ ) + 1 ) , if 0 ≤ k < m + ℓ , if k ≥ m + ℓ .
Proof. Since f f f and g g g are reduced and coprime, f g fg f g is reduced, so it has an isolated singularity at the origin and μ ( f g ) < ∞ \mu(fg)<\infty μ ( f g ) < ∞ . Since O 2 \mathcal{O}_2 O 2 is a domain, ν ( f g ) = ν ( f ) + ν ( g ) = m + ℓ \nu(fg)=\nu(f)+\nu(g)=m+\ell ν ( f g ) = ν ( f ) + ν ( g ) = m + ℓ . Now apply Corollary 2.1 to f g fg f g . □
On a conjecture of Hussain–Yau–Zuo
In [1 ] , P. Almirón studied the quotient between the Milnor number and the Tjurina number of an isolated complete intersection singularity and established its connections with other problems in singularity theory. As part of this study, he proposed the following conjecture [1 Conjecture 4.10] .
Conjecture 3.1. Let f ∈ m ⊂ O 3 f\in\mathfrak{m}\subset\mathcal{O}_3 f ∈ m ⊂ O 3 define an isolated hypersurface singularity ( V , 0 ) ⊂ ( C 3 , 0 ) (V,0)\subset(\mathbb{C}^3,0) ( V , 0 ) ⊂ ( C 3 , 0 ) . Then
μ ( f ) τ ( f ) < 3 2 . \frac{\mu(f)}{\tau(f)}<\frac{3}{2}. τ ( f ) μ ( f ) < 2 3 .
Almirón showed that Conjecture 3.1 follows from Durfee’s conjecture 6 p g ≤ μ 6p_g\leq\mu 6 p g ≤ μ (here p g p_g p g denotes the geometric genus of an isolated surface singularity), and verified it for several families of surface singularities for which Durfee’s conjecture is known to hold. Recently, Ma–Zuo [17 ] proved Conjecture 3.1 in full generality, without relying on Durfee’s conjecture, as a consequence of a sharp upper bound for μ / τ \mu/\tau μ / τ valid for isolated hypersurface singularities in any dimension [17 Theorem 4.5 and Corollary 4.8] .
In the context of the k k k -th Milnor and Tjurina numbers of hypersurface singularities, Hussain–Yau–Zuo [14 Conjecture 1.4] conjectured the following inequality.
Conjecture 3.2. Let f ∈ m ⊂ O 3 f\in\mathfrak{m}\subset\mathcal{O}_3 f ∈ m ⊂ O 3 define an isolated hypersurface singularity ( V , 0 ) ⊂ ( C 3 , 0 ) (V,0)\subset(\mathbb{C}^3,0) ( V , 0 ) ⊂ ( C 3 , 0 ) . Then for each integer k ≥ 0 k\geq0 k ≥ 0 ,
μ k ( f ) τ k ( f ) < 3 2 . \frac{\mu^k(f)}{\tau^k(f)}<\frac{3}{2}. τ k ( f ) μ k ( f ) < 2 3 .
Since μ 0 ( f ) = μ ( f ) \mu^0(f)=\mu(f) μ 0 ( f ) = μ ( f ) and τ 0 ( f ) = τ ( f ) \tau^0(f)=\tau(f) τ 0 ( f ) = τ ( f ) , the case k = 0 k=0 k = 0 of Conjecture 3.2 is precisely Conjecture 3.1 , and hence it holds by [17 Corollary 4.8] . The question is therefore whether the bound 3 / 2 3/2 3/2 persists for k ≥ 1 k\geq1 k ≥ 1 . Hussain–Yau–Zuo considered the hypersurfaces defined by f 1 = x 3 m + 1 + y 3 m + 1 + z 3 m + 1 + x m + 1 y m + 1 z m + 1 f_1=x^{3m+1}+y^{3m+1}+z^{3m+1}+x^{m+1}y^{m+1}z^{m+1} f 1 = x 3 m + 1 + y 3 m + 1 + z 3 m + 1 + x m + 1 y m + 1 z m + 1 and f 2 = x 3 m − 1 + y 3 m − 1 + z 3 m − 1 + x m y m z m f_2=x^{3m-1}+y^{3m-1}+z^{3m-1}+x^my^mz^m f 2 = x 3 m − 1 + y 3 m − 1 + z 3 m − 1 + x m y m z m , with m ≥ 1 m\geq1 m ≥ 1 , and verified explicitly that μ k / τ k < 3 / 2 \mu^k/\tau^k<3/2 μ k / τ k < 3/2 for k = 0 , 1 , 2 k=0,1,2 k = 0 , 1 , 2 . We show that Conjecture 3.2 fails already for the simplest isolated hypersurface singularity.
The computation of the k k k -th Tjurina numbers below rests on the following lemma, which gives dim C O n / ( ( f ) + m N ) \dim_{\mathbb C}\mathcal O_n/((f)+\mathfrak m^N) dim C O n / (( f ) + m N ) . For a nonzero germ h ∈ O 2 h\in\mathcal O_2 h ∈ O 2 , we denote by in ( h ) \operatorname{in}(h) in ( h ) its initial form , that is, the nonzero homogeneous component of lowest degree in its Taylor expansion. Thus deg in ( h ) = ν ( h ) \deg\operatorname{in}(h)=\nu(h) deg in ( h ) = ν ( h ) .
Lemma 3.3. Let f ∈ m ⊂ O n f\in\mathfrak m\subset\mathcal O_n f ∈ m ⊂ O n , let ν = ν ( f ) \nu=\nu(f) ν = ν ( f ) and let N ≥ 0 N\geq0 N ≥ 0 be an integer. Put m j = O n \mathfrak m^j=\mathcal O_n m j = O n for j ≤ 0 j\leq0 j ≤ 0 . Then dim C O n / m j = ( j + n − 1 n ) \dim_{\mathbb C}\mathcal O_n/\mathfrak m^j=\binom{j+n-1}{n} dim C O n / m j = ( n j + n − 1 ) for every j ≥ 0 j\geq0 j ≥ 0 , ( m N : f ) = m N − ν (\mathfrak m^N:f)=\mathfrak m^{N-\nu} ( m N : f ) = m N − ν and
dim C O n ( f ) + m N = dim C O n m N − dim C O n m N − ν . \dim_{\mathbb C}\frac{\mathcal O_n}{(f)+\mathfrak m^N}=\dim_{\mathbb C}\frac{\mathcal O_n}{\mathfrak m^N}-\dim_{\mathbb C}\frac{\mathcal O_n}{\mathfrak m^{N-\nu}}. dim C ( f ) + m N O n = dim C m N O n − dim C m N − ν O n . In particular, for N ≥ ν N\geq\nu N ≥ ν this equals ( N + n − 1 n ) − ( N − ν + n − 1 n ) \binom{N+n-1}{n}-\binom{N-\nu+n-1}{n} ( n N + n − 1 ) − ( n N − ν + n − 1 ) .
Proof. For j ≥ 0 j\geq0 j ≥ 0 , the monomials of degree less than j j j form a basis of O n / m j \mathcal O_n/\mathfrak m^j O n / m j , and there are ( d + n − 1 n − 1 ) \binom{d+n-1}{n-1} ( n − 1 d + n − 1 ) monomials of degree d d d ; hence dim O n / m j = ∑ d = 0 j − 1 ( d + n − 1 n − 1 ) = ( j + n − 1 n ) \dim\mathcal O_n/\mathfrak m^j=\sum_{d=0}^{j-1}\binom{d+n-1}{n-1}=\binom{j+n-1}{n} dim O n / m j = ∑ d = 0 j − 1 ( n − 1 d + n − 1 ) = ( n j + n − 1 ) .
Let φ : O n → O n / m N \varphi:\mathcal O_n\to\mathcal O_n/\mathfrak m^N φ : O n → O n / m N , φ ( h ) = f h + m N \varphi(h)=fh+\mathfrak m^N φ ( h ) = f h + m N . It is C \mathbb C C -linear and its image is ( ( f ) + m N ) / m N ((f)+\mathfrak m^N)/\mathfrak m^N (( f ) + m N ) / m N . Its kernel is ( m N : f ) = { h : f h ∈ m N } (\mathfrak m^N:f)=\{h:fh\in\mathfrak m^N\} ( m N : f ) = { h : f h ∈ m N } , so by the first isomorphism theorem the image is isomorphic to O n / ( m N : f ) \mathcal O_n/(\mathfrak m^N:f) O n / ( m N : f ) . The initial form of f h fh f h is the product of the initial forms of f f f and h h h , which is nonzero; hence ν ( f h ) = ν + ν ( h ) \nu(fh)=\nu+\nu(h) ν ( f h ) = ν + ν ( h ) . Then f h ∈ m N fh\in\mathfrak m^N f h ∈ m N if and only if ν ( h ) ≥ N − ν \nu(h)\geq N-\nu ν ( h ) ≥ N − ν , that is, ( m N : f ) = m N − ν (\mathfrak m^N:f)=\mathfrak m^{N-\nu} ( m N : f ) = m N − ν . Finally, let ι \iota ι be the inclusion ( ( f ) + m N ) / m N ↪ O n / m N ((f)+\mathfrak m^N)/\mathfrak m^N\hookrightarrow\mathcal O_n/\mathfrak m^N (( f ) + m N ) / m N ↪ O n / m N and let π \pi π be the canonical projection O n / m N → O n / ( ( f ) + m N ) \mathcal O_n/\mathfrak m^N\to\mathcal O_n/((f)+\mathfrak m^N) O n / m N → O n / (( f ) + m N ) , which is well defined since m N ⊂ ( f ) + m N \mathfrak m^N\subset(f)+\mathfrak m^N m N ⊂ ( f ) + m N . Then ι \iota ι is injective, π \pi π is surjective and ker π = im ι \ker\pi=\operatorname{im}\iota ker π = im ι , so the sequence 0 → ( ( f ) + m N ) / m N → ι O n / m N → π O n / ( ( f ) + m N ) → 0 0\to((f)+\mathfrak m^N)/\mathfrak m^N\xrightarrow{\iota}\mathcal O_n/\mathfrak m^N\xrightarrow{\pi}\mathcal O_n/((f)+\mathfrak m^N)\to0 0 → (( f ) + m N ) / m N ι O n / m N π O n / (( f ) + m N ) → 0 is exact. It gives dim O n / ( ( f ) + m N ) = dim O n / m N − dim O n / ( m N : f ) \dim\mathcal O_n/((f)+\mathfrak m^N)=\dim\mathcal O_n/\mathfrak m^N-\dim\mathcal O_n/(\mathfrak m^N:f) dim O n / (( f ) + m N ) = dim O n / m N − dim O n / ( m N : f ) , and the stated formulas follow. □
Proposition 3.4. Let f = x 2 + y 2 + z 2 ∈ C { x , y , z } f=x^2+y^2+z^2\in\mathbb C\{x,y,z\} f = x 2 + y 2 + z 2 ∈ C { x , y , z } . Then for every integer k ≥ 0 k\geq0 k ≥ 0
μ k ( f ) = ( k + 3 3 ) , τ k ( f ) = ( k + 1 ) 2 , \mu^k(f)=\binom{k+3}{3},\qquad\tau^k(f)=(k+1)^2, μ k ( f ) = ( 3 k + 3 ) , τ k ( f ) = ( k + 1 ) 2 , and μ k ( f ) / τ k ( f ) > 3 / 2 \mu^k(f)/\tau^k(f)>3/2 μ k ( f ) / τ k ( f ) > 3/2 if and only if k ≥ 5 k\geq5 k ≥ 5 . In particular, μ 5 ( f ) = 56 \mu^5(f)=56 μ 5 ( f ) = 56 , τ 5 ( f ) = 36 \tau^5(f)=36 τ 5 ( f ) = 36 and μ 5 ( f ) / τ 5 ( f ) = 14 / 9 > 3 / 2 \mu^5(f)/\tau^5(f)=14/9>3/2 μ 5 ( f ) / τ 5 ( f ) = 14/9 > 3/2 .
Proof. Since J ( f ) = ( 2 x , 2 y , 2 z ) = m J(f)=(2x,2y,2z)=\mathfrak m J ( f ) = ( 2 x , 2 y , 2 z ) = m , we have m k J ( f ) = m k + 1 \mathfrak m^kJ(f)=\mathfrak m^{k+1} m k J ( f ) = m k + 1 . Hence μ k ( f ) = dim C O 3 / m k + 1 = ( k + 3 3 ) \mu^k(f)=\dim_{\mathbb C}\mathcal O_3/\mathfrak m^{k+1}=\binom{k+3}{3} μ k ( f ) = dim C O 3 / m k + 1 = ( 3 k + 3 ) by the formula for dim C O n / m j \dim_{\mathbb C}\mathcal O_n/\mathfrak m^j dim C O n / m j in Lemma 3.3 with n = 3 n=3 n = 3 , j = k + 1 j=k+1 j = k + 1 .
Put N = k + 1 N=k+1 N = k + 1 . Since ν ( f ) = 2 \nu(f)=2 ν ( f ) = 2 , Lemma 3.3 gives
τ k ( f ) = dim C O 3 m k + 1 − dim C O 3 m k − 1 = ( k + 3 3 ) − ( k + 1 3 ) . \tau^k(f)=\dim_{\mathbb C}\frac{\mathcal O_3}{\mathfrak m^{k+1}}-\dim_{\mathbb C}\frac{\mathcal O_3}{\mathfrak m^{k-1}}=\binom{k+3}{3}-\binom{k+1}{3}. τ k ( f ) = dim C m k + 1 O 3 − dim C m k − 1 O 3 = ( 3 k + 3 ) − ( 3 k + 1 ) .
(for k = 0 k=0 k = 0 we have m − 1 = O 3 \mathfrak m^{-1}=\mathcal O_3 m − 1 = O 3 and ( 1 3 ) = 0 \binom{1}{3}=0 ( 3 1 ) = 0 ). Since ( k + 3 ) ( k + 2 ) ( k + 1 ) − ( k + 1 ) k ( k − 1 ) = 6 ( k + 1 ) 2 (k+3)(k+2)(k+1)-(k+1)k(k-1)=6(k+1)^2 ( k + 3 ) ( k + 2 ) ( k + 1 ) − ( k + 1 ) k ( k − 1 ) = 6 ( k + 1 ) 2 , we get τ k ( f ) = ( k + 1 ) 2 \tau^k(f)=(k+1)^2 τ k ( f ) = ( k + 1 ) 2 .
Finally, μ k ( f ) / τ k ( f ) > 3 / 2 \mu^k(f)/\tau^k(f)>3/2 μ k ( f ) / τ k ( f ) > 3/2 is equivalent to 2 μ k ( f ) > 3 τ k ( f ) 2\mu^k(f)>3\tau^k(f) 2 μ k ( f ) > 3 τ k ( f ) , which can be rewritten as ( k + 3 ) ( k + 2 ) ( k + 1 ) > 9 ( k + 1 ) 2 (k+3)(k+2)(k+1)>9(k+1)^2 ( k + 3 ) ( k + 2 ) ( k + 1 ) > 9 ( k + 1 ) 2 . Dividing by k + 1 > 0 k+1>0 k + 1 > 0 , this becomes ( k + 3 ) ( k + 2 ) > 9 ( k + 1 ) (k+3)(k+2)>9(k+1) ( k + 3 ) ( k + 2 ) > 9 ( k + 1 ) , that is, to k 2 − 4 k − 3 > 0 k^2-4k-3>0 k 2 − 4 k − 3 > 0 . The roots of this polynomial are 2 ± 7 2\pm\sqrt{7} 2 ± 7 and its leading coefficient is positive, so for k ≥ 0 k \ge0 k ≥ 0 the inequality holds exactly when k > 2 + 7 k > 2 + \sqrt{7} k > 2 + 7 , that is, for the integers k ≥ 5 k \ge5 k ≥ 5 , since 2 < 7 < 3 2 < \sqrt{7} < 3 2 < 7 < 3 .
Theorem 3.5. Let n ≥ 2 n \ge2 n ≥ 2 and let f ∈ m ⊂ O n = C { x 1 , … , x n } f \in\mathfrak{m} \subset\mathcal{O}_n = \mathbb{C}\{x_1,\ldots,x_n\} f ∈ m ⊂ O n = C { x 1 , … , x n } with μ ( f ) < ∞ \mu(f) < \infty μ ( f ) < ∞ . Then lim k → ∞ μ k ( f ) / τ k ( f ) = ∞ \lim_{k\to\infty} \mu^k(f)/\tau^k(f) = \infty lim k → ∞ μ k ( f ) / τ k ( f ) = ∞ . In particular, for n = 3 n = 3 n = 3 Conjecture 3.2 fails for every such f f f and every k k k large enough.
Proof. Let ν = ν ( f ) \nu= \nu(f) ν = ν ( f ) and r = μ ( f ) r = \mu(f) r = μ ( f ) . If ν ( f ) = 1 \nu(f) = 1 ν ( f ) = 1 , a biholomorphic change of coordinates reduces to f = x 1 f = x_1 f = x 1 . For every k ≥ 1 k \ge1 k ≥ 1 , we then have
μ k ( f ) = ( k + n − 1 n ) , τ k ( f ) = ( k + n − 2 n − 1 ) . \mu^k(f) = \binom{k+n-1}{n}, \qquad\tau^k(f) = \binom{k+n-2}{n-1}. μ k ( f ) = ( n k + n − 1 ) , τ k ( f ) = ( n − 1 k + n − 2 ) .
and hence
μ k ( f ) τ k ( f ) = k + n − 1 n ⟶ ∞ . \frac{\mu^k(f)}{\tau^k(f)} = \frac{k+n-1}{n} \longrightarrow\infty. τ k ( f ) μ k ( f ) = n k + n − 1 ⟶ ∞.
We may therefore assume that ν ( f ) ≥ 2 \nu(f) \ge2 ν ( f ) ≥ 2 . Write f = ∑ d ≥ ν f d f = \sum_{d\ge\nu} f_d f = ∑ d ≥ ν f d with f d f_d f d homogeneous of degree d d d . Then ∂ f / ∂ x i = ∑ d ≥ ν ∂ f d / ∂ x i \partial f/\partial x_i = \sum_{d\ge\nu} \partial f_d/\partial x_i ∂ f / ∂ x i = ∑ d ≥ ν ∂ f d / ∂ x i , and each ∂ f d / ∂ x i \partial f_d/\partial x_i ∂ f d / ∂ x i is zero or homogeneous of degree d − 1 ≥ ν − 1 d-1 \ge\nu-1 d − 1 ≥ ν − 1 . Hence ∂ f / ∂ x i ∈ m ν − 1 \partial f/\partial x_i \in\mathfrak{m}^{\nu-1} ∂ f / ∂ x i ∈ m ν − 1 for every i i i , so J ( f ) ⊂ m ν − 1 J(f) \subset\mathfrak{m}^{\nu-1} J ( f ) ⊂ m ν − 1 and m k J ( f ) ⊂ m k + ν − 1 \mathfrak{m}^kJ(f) \subset\mathfrak{m}^{k+\nu-1} m k J ( f ) ⊂ m k + ν − 1 . Therefore μ k ( f ) = dim C O n / m k J ( f ) ≥ dim C O n / m k + ν − 1 \mu^k(f) = \dim_{\mathbb{C}}\mathcal{O}_n/\mathfrak{m}^kJ(f) \ge\dim_{\mathbb{C}}\mathcal{O}_n/\mathfrak{m}^{k+\nu-1} μ k ( f ) = dim C O n / m k J ( f ) ≥ dim C O n / m k + ν − 1 . By the formula for dim C O n / m j \dim_{\mathbb{C}}\mathcal{O}_n/\mathfrak{m}^j dim C O n / m j in Lemma 3.3 with j = k + ν − 1 ≥ 0 j = k+\nu-1 \ge0 j = k + ν − 1 ≥ 0 ,
dim C O n / m k + ν − 1 = ( ( k + ν − 1 ) + n − 1 n ) = ( k + ν + n − 2 n ) . \dim_{\mathbb{C}}\mathcal{O}_n/\mathfrak{m}^{k+\nu-1}
= \binom{(k+\nu-1)+n-1}{n}
= \binom{k+\nu+n-2}{n}. dim C O n / m k + ν − 1 = ( n ( k + ν − 1 ) + n − 1 ) = ( n k + ν + n − 2 ) .
Recall that r = μ ( f ) = dim C O n / J ( f ) r = \mu(f) = \dim_{\mathbb{C}}\mathcal{O}_n/J(f) r = μ ( f ) = dim C O n / J ( f ) . The algebra A = O n / J ( f ) A = \mathcal{O}_n/J(f) A = O n / J ( f ) is local, with maximal ideal m ‾ = m / J ( f ) \overline{\mathfrak{m}} = \mathfrak{m}/J(f) m = m / J ( f ) . If m ‾ i = m ‾ i + 1 \overline{\mathfrak{m}}^i = \overline{\mathfrak{m}}^{i+1} m i = m i + 1 for some i i i , then m ‾ i = 0 \overline{\mathfrak{m}}^i = 0 m i = 0 by Nakayama’s lemma, since m ‾ i \overline{\mathfrak{m}}^i m i is a finitely generated A A A -module and m ‾ ⋅ m ‾ i = m ‾ i \overline{\mathfrak{m}}\cdot\overline{\mathfrak{m}}^i = \overline{\mathfrak{m}}^i m ⋅ m i = m i . Hence, as long as m ‾ i ≠ 0 \overline{\mathfrak{m}}^i \ne0 m i = 0 , the inclusion m ‾ i + 1 ⊂ m ‾ i \overline{\mathfrak{m}}^{i+1} \subset\overline{\mathfrak{m}}^i m i + 1 ⊂ m i is strict, so it loses at least one dimension. Since dim C A = r \dim_{\mathbb{C}} A = r dim C A = r , this gives dim C m ‾ i ≤ r − i \dim_{\mathbb{C}}\overline{\mathfrak{m}}^i \le r-i dim C m i ≤ r − i and, in particular, m ‾ r = 0 \overline{\mathfrak{m}}^r = 0 m r = 0 , that is, m r ⊂ J ( f ) \mathfrak{m}^r \subset J(f) m r ⊂ J ( f ) . Therefore m k + r ⊂ m k J ( f ) \mathfrak{m}^{k+r} \subset\mathfrak{m}^kJ(f) m k + r ⊂ m k J ( f ) , so ( f ) + m k + r ⊂ ( f ) + m k J ( f ) (f) + \mathfrak{m}^{k+r} \subset(f) + \mathfrak{m}^kJ(f) ( f ) + m k + r ⊂ ( f ) + m k J ( f ) and, for N = k + r N = k+r N = k + r , τ k ( f ) ≤ dim C O n / ( ( f ) + m N ) \tau^k(f) \le\dim_{\mathbb{C}}\mathcal{O}_n/((f) + \mathfrak{m}^N) τ k ( f ) ≤ dim C O n / (( f ) + m N ) . Let k ≥ ν k \ge\nu k ≥ ν , so that N ≥ ν N \ge\nu N ≥ ν . By Lemma 3.3 ,
dim C O n ( f ) + m N = ( N + n − 1 n ) − ( N − ν + n − 1 n ) = ∑ i = 0 ν − 1 ( N + n − 2 − i n − 1 ) ≤ ν ( N + n − 2 n − 1 ) . \dim_{\mathbb{C}}\frac{\mathcal{O}_n}{(f)+\mathfrak{m}^N}
= \binom{N+n-1}{n} - \binom{N-\nu+n-1}{n}
= \sum_{i=0}^{\nu-1}\binom{N+n-2-i}{n-1}
\le\nu\binom{N+n-2}{n-1}. dim C ( f ) + m N O n = ( n N + n − 1 ) − ( n N − ν + n − 1 ) = i = 0 ∑ ν − 1 ( n − 1 N + n − 2 − i ) ≤ ν ( n − 1 N + n − 2 ) .
where the equality of sums follows from Pascal’s rule ( a n ) − ( a − 1 n ) = ( a − 1 n − 1 ) \binom{a}{n} - \binom{a-1}{n} = \binom{a-1}{n-1} ( n a ) − ( n a − 1 ) = ( n − 1 a − 1 ) applied ν \nu ν times. Thus, for k ≥ ν k \ge\nu k ≥ ν ,
μ k ( f ) τ k ( f ) ≥ ( k + ν + n − 2 n ) ν ( k + r + n − 2 n − 1 ) . \frac{\mu^k(f)}{\tau^k(f)}
\ge\frac{\binom{k+\nu+n-2}{n}}{\nu\binom{k+r+n-2}{n-1}}. τ k ( f ) μ k ( f ) ≥ ν ( n − 1 k + r + n − 2 ) ( n k + ν + n − 2 ) .
The numerator is a polynomial of degree n n n in k k k and the denominator a polynomial of degree n − 1 n-1 n − 1 , so the right-hand side tends to ∞ \infty ∞ . □
Remark 3.6. The plane curve conjecture was already disproved in [21 Section 7] . The node x 2 + y 2 x^2+y^2 x 2 + y 2 provides the following simpler counterexample. The same proof as in Proposition 3.4 , now in two variables (Lemma 3.3 with n = 2 n=2 n = 2 ), gives μ k ( x 2 + y 2 ) = ( k + 2 2 ) \mu^k(x^2+y^2)=\binom{k+2}{2} μ k ( x 2 + y 2 ) = ( 2 k + 2 ) and τ k ( x 2 + y 2 ) = 2 k + 1 \tau^k(x^2+y^2)=2k+1 τ k ( x 2 + y 2 ) = 2 k + 1 , so μ k / τ k > 4 / 3 \mu^k/\tau^k>4/3 μ k / τ k > 4/3 if and only if 3 k 2 − 7 k − 2 > 0 3k^2-7k-2>0 3 k 2 − 7 k − 2 > 0 , that is, k ≥ 3 k\geq3 k ≥ 3 (for instance μ 3 = 10 \mu^3=10 μ 3 = 10 , τ 3 = 7 \tau^3=7 τ 3 = 7 ). Hence the analogous conjecture for plane curves [14 Conjecture 1.3] also fails.
Hussain–Yau–Zuo supported Conjecture 3.2 with the families f 1 = x 3 m + 1 + y 3 m + 1 + z 3 m + 1 + x m + 1 y m + 1 z m + 1 f_1=x^{3m+1}+y^{3m+1}+z^{3m+1}+x^{m+1}y^{m+1}z^{m+1} f 1 = x 3 m + 1 + y 3 m + 1 + z 3 m + 1 + x m + 1 y m + 1 z m + 1 and f 2 = x 3 m − 1 + y 3 m − 1 + z 3 m − 1 + x m y m z m f_2=x^{3m-1}+y^{3m-1}+z^{3m-1}+x^my^mz^m f 2 = x 3 m − 1 + y 3 m − 1 + z 3 m − 1 + x m y m z m , with m ≥ 1 m\geq1 m ≥ 1 : they computed μ k / τ k \mu^k/\tau^k μ k / τ k only for k = 0 , 1 , 2 k=0,1,2 k = 0 , 1 , 2 , obtained values close to 27 / 19 < 3 / 2 27/19<3/2 27/19 < 3/2 for m m m large, and wrote that they had no example with k ≥ 1 k\geq1 k ≥ 1 and quotient close to 3 / 2 3/2 3/2 [14 ] . Proposition 3.4 and Theorem 3.5 show that checking only k ≤ 2 k\leq2 k ≤ 2 cannot detect the failure of Conjecture 3.2 : for x 2 + y 2 + z 2 x^2+y^2+z^2 x 2 + y 2 + z 2 the quotient exceeds 3 / 2 3/2 3/2 only from k = 5 k=5 k = 5 on, and in general it tends to infinity with k k k . Theorem 3.5 applies to f 1 f_1 f 1 and f 2 f_2 f 2 , which have isolated singularities, as required in Conjecture 3.2 . Hence μ k ( f i ) / τ k ( f i ) → ∞ \mu^k(f_i)/\tau^k(f_i)\to\infty μ k ( f i ) / τ k ( f i ) → ∞ as k → ∞ k\to\infty k → ∞ for every m ≥ 1 m\geq1 m ≥ 1 . For instance, for m = 2 m=2 m = 2 and f 1 = x 7 + y 7 + z 7 + x 3 y 3 z 3 f_1=x^7+y^7+z^7+x^3y^3z^3 f 1 = x 7 + y 7 + z 7 + x 3 y 3 z 3 , computations with SINGULAR [5 ] give μ 16 ( f 1 ) = 2024 \mu^{16}(f_1)=2024 μ 16 ( f 1 ) = 2024 and τ 16 ( f 1 ) = 1344 \tau^{16}(f_1)=1344 τ 16 ( f 1 ) = 1344 , so μ 16 ( f 1 ) / τ 16 ( f 1 ) = 253 / 168 > 3 / 2 \mu^{16}(f_1)/\tau^{16}(f_1)=253/168>3/2 μ 16 ( f 1 ) / τ 16 ( f 1 ) = 253/168 > 3/2 .
On the k k k -th Milnor number of a semiquasihomogeneous function
A polynomial f ∈ C [ x 1 , … , x n ] f\in\mathbb{C}[x_1,\ldots,x_n] f ∈ C [ x 1 , … , x n ] is called quasihomogeneous of degree d d d if there exists a vector w = ( w 1 , … , w n ) ∈ N > 0 n w=(w_1,\ldots,w_n)\in\mathbb{N}_{>0}^n w = ( w 1 , … , w n ) ∈ N > 0 n and d ∈ N d\in\mathbb{N} d ∈ N such that every monomial x α x^\alpha x α of f f f with nonzero coefficient satisfies ∑ i = 1 n α i w i = d \sum_{i=1}^n\alpha_iw_i=d ∑ i = 1 n α i w i = d . In this case we say that f f f is of type ( w 1 , … , w n , d ) (w_1,\ldots,w_n,d) ( w 1 , … , w n , d ) . We call w 1 , … , w n w_1,\ldots,w_n w 1 , … , w n the weights of the variables and d d d the weighted degree of f f f , denoted as deg w ( f ) = d \deg_w(f)=d deg w ( f ) = d . A quasihomogeneous polynomial f f f is said to be nondegenerate if 0 0 0 is an isolated critical point. A power series f ∈ O n f\in\mathcal{O}_n f ∈ O n is said to be semiquasihomogeneous (SQH) of degree d d d with weights w 1 , … , w n w_1,\ldots,w_n w 1 , … , w n if it is of the form f = f w + f ′ f=f_w+f' f = f w + f ′ , where f w f_w f w is a nondegenerate quasihomogeneous polynomial of degree d d d with weights w i w_i w i and f ′ f' f ′ is a power series all of whose monomials have weighted degree strictly greater than d d d .
Let K [ [ x 1 , … , x n ] ] K[[x_1,\ldots,x_n]] K [[ x 1 , … , x n ]] be the ring of formal power series over a field K K K . For brevity, we write K [ [ x 1 , … , x n ] ] K[[x_1,\ldots,x_n]] K [[ x 1 , … , x n ]] as K [ [ x ] ] K[[x]] K [[ x ]] . Let m \mathfrak{m} m be the maximal ideal of K [ [ x ] ] K[[x]] K [[ x ]] . In this situation, we can also define the k k k -th Milnor number of f ∈ K [ [ x ] ] f\in K[[x]] f ∈ K [[ x ]] as μ k ( f ) : = dim K K [ [ x ] ] / m k J ( f ) \mu^k(f):=\dim_K K[[x]]/\mathfrak{m}^kJ(f) μ k ( f ) := dim K K [[ x ]] / m k J ( f ) , where J ( f ) J(f) J ( f ) is the Jacobian ideal of f f f in K [ [ x ] ] K[[x]] K [[ x ]] . If the dimension is infinite, we simply write μ k ( f ) = ∞ \mu^k(f)=\infty μ k ( f ) = ∞ . In what follows in this section K = C K=\mathbb{C} K = C .
For a SQH power series f f f of weighted degree d d d with weights w 1 , … , w n w_1,\ldots,w_n w 1 , … , w n we have μ ( f ) = μ ( f w ) < ∞ \mu(f)=\mu(f_w)<\infty μ ( f ) = μ ( f w ) < ∞ . This is proved in [2 p. 194] for convergent power series, by a topological argument (the degree of the gradient map) which does not apply to formal series. For formal power series, [4 Proposition 2] gives, by an algebraic argument, μ ( f ) = ∏ i = 1 n ( d / w i − 1 ) \mu(f)=\prod_{i=1}^n(d/w_i-1) μ ( f ) = ∏ i = 1 n ( d / w i − 1 ) , which depends only on d d d and w w w ; applying it to f f f and to f w f_w f w gives μ ( f ) = μ ( f w ) \mu(f)=\mu(f_w) μ ( f ) = μ ( f w ) . Hence m N ⊂ J ( f ) \mathfrak{m}^N\subset J(f) m N ⊂ J ( f ) for some N N N and m k J ( f ) ⊃ m k + N \mathfrak{m}^kJ(f)\supset\mathfrak{m}^{k+N} m k J ( f ) ⊃ m k + N . Since C { x } / m s = C [ [ x ] ] / m s \mathbb{C}\{x\}/\mathfrak{m}^s=\mathbb{C}[[x]]/\mathfrak{m}^s C { x } / m s = C [[ x ]] / m s for every s ≥ 0 s\geq0 s ≥ 0 , the number μ k ( f ) \mu^k(f) μ k ( f ) is the same whether computed in C { x } \mathbb{C}\{x\} C { x } or in C [ [ x ] ] \mathbb{C}[[x]] C [[ x ]] .
In [23 Conjecture 4.65] , Shi–Yau–Zuo proposed the following conjecture.
Conjecture 4.1. Let f = f w + f ′ ∈ C [ [ x ] ] f=f_w+f^\prime\in\mathbb{C}[[x]] f = f w + f ′ ∈ C [[ x ]] be a semiquasihomogeneous (SQH) power series with respect to a weight vector w ∈ N > 0 w\in\mathbb{N}_{>0} w ∈ N > 0 , and assume that μ ( f w ) < ∞ \mu(f_w)<\infty μ ( f w ) < ∞ . Then μ k ( f w ) = μ k ( f ) \mu^k(f_w)=\mu^k(f) μ k ( f w ) = μ k ( f ) for every integer k ≥ 0 k\geq0 k ≥ 0 .
In [23 Theorem C] , the authors proved Conjecture 4.1 under the additional assumption that k ≤ min i { ν ( f x i ) } k\leq\min_i\{\nu(f_{x_i})\} k ≤ min i { ν ( f x i )} .
Liu–Yau–Zhang [16 Main Theorem A] recently proved Conjecture 4.1 for convergent germs in the equal-weight case, for every k ≥ 0 k\geq0 k ≥ 0 and every n ≥ 2 n\geq2 n ≥ 2 . More precisely, their result applies to germs f = f r + f r + 1 + ⋯ f=f_r+f_{r+1}+\cdots f = f r + f r + 1 + ⋯ with r ≥ 2 r\geq2 r ≥ 2 , where each f j f_j f j is homogeneous of degree j j j and V + ( f r ) ⊂ P n − 1 V_+(f_r)\subset\mathbb{P}^{n-1} V + ( f r ) ⊂ P n − 1 is smooth (equivalently, f r f_r f r has an isolated critical point at the origin).
In this section, we show that the conjecture fails in dimension three for unequal weights. Our counterexample lies outside the range k ≤ min i { ν ( f x i ) } k\leq\min_i\{\nu(f_{x_i})\} k ≤ min i { ν ( f x i )} established in [23 Theorem C] . We also prove that the conjecture holds in dimension two over C \mathbb{C} C for arbitrary positive weights and every k ≥ 0 k\geq0 k ≥ 0 ; see Theorem 4.2 .
To this end, consider the germ
f = x 3 + y 7 + z 7 + x y 2 z 3 f=x^3+y^7+z^7+xy^2z^3 f = x 3 + y 7 + z 7 + x y 2 z 3
in O 3 = C { x , y , z } \mathcal{O}_3=\mathbb{C}\{x,y,z\} O 3 = C { x , y , z } . Its principal part with respect to the weight vector w = ( 7 , 3 , 3 ) w=(7,3,3) w = ( 7 , 3 , 3 ) is
f w = x 3 + y 7 + z 7 . f_w=x^3+y^7+z^7. f w = x 3 + y 7 + z 7 .
Since f w f_w f w has weighted degree 21, the monomial x y 2 z 3 xy^2z^3 x y 2 z 3 has weighted degree 22, and μ ( f w ) = 72 < ∞ \mu(f_w)=72<\infty μ ( f w ) = 72 < ∞ , the germ f f f is semiquasihomogeneous. Using Singular [5 ] with a local monomial ordering, we compute the dimensions of O 3 / J ( h ) \mathcal{O}_3/J(h) O 3 / J ( h ) and O 3 / m 6 J ( h ) \mathcal{O}_3/\mathfrak{m}^6J(h) O 3 / m 6 J ( h ) for h = f w h=f_w h = f w and h = f h=f h = f , and we obtain μ ( f w ) = μ ( f ) = 72 \mu(f_w)=\mu(f)=72 μ ( f w ) = μ ( f ) = 72 and μ 6 ( f w ) = 200 ≠ 199 = μ 6 ( f ) \mu^6(f_w)=200\ne199=\mu^6(f) μ 6 ( f w ) = 200 = 199 = μ 6 ( f ) , which gives a negative answer to Conjecture 4.1 in dimension three.
Note that ν ( f x ) = 2 \nu(f_x)=2 ν ( f x ) = 2 and ν ( f y ) = ν ( f z ) = 5 \nu(f_y)=\nu(f_z)=5 ν ( f y ) = ν ( f z ) = 5 , so k = 6 > min { ν ( f x ) , ν ( f y ) , ν ( f z ) } = 2 k=6>\min\{\nu(f_x),\nu(f_y),\nu(f_z)\}=2 k = 6 > min { ν ( f x ) , ν ( f y ) , ν ( f z )} = 2 , which is consistent with [23 Theorem C] .
We now prove that the conjecture holds in dimension two over the ring C [ [ x , y ] ] \mathbb{C}[[x,y]] C [[ x , y ]] .
In the proof of the next theorem we use [21 Corollary 4.1] , which gives the formula for μ k ( f ) \mu^k(f) μ k ( f ) of a plane curve. It is stated for convergent power series, but its proof uses only algebraic arguments (orders of products, relatively prime elements of a unique factorization domain, and exact sequences of vector spaces), which are also valid in C [ [ x , y ] ] \mathbb{C}[[x,y]] C [[ x , y ]] . Hence the formula holds for every f ∈ C [ [ x , y ] ] f\in\mathbb{C}[[x,y]] f ∈ C [[ x , y ]] with μ ( f ) < ∞ \mu(f)<\infty μ ( f ) < ∞ .
Theorem 4.2. Suppose that f = f w + f ′ ∈ C [ [ x , y ] ] f=f_w+f^\prime\in\mathbb{C}[[x,y]] f = f w + f ′ ∈ C [[ x , y ]] is a semiquasihomogeneous (SQH) power series with respect to a weight vector w ∈ N > 0 2 w\in\mathbb{N}_{>0}^2 w ∈ N > 0 2 . Then μ k ( f w ) = μ k ( f ) \mu^k(f_w)=\mu^k(f) μ k ( f w ) = μ k ( f ) for every integer k ≥ 0 k\geq0 k ≥ 0 .
Proof. By [4 Proposition 2] , applied to f f f and to f w f_w f w , which are SQH with respect to the same w w w and the same degree d d d , μ ( f ) = ∏ i = 1 2 ( d / w i − 1 ) = μ ( f w ) < ∞ \mu(f) = \prod_{i=1}^{2}(d/w_i - 1) = \mu(f_w) < \infty μ ( f ) = ∏ i = 1 2 ( d / w i − 1 ) = μ ( f w ) < ∞ . Hence f f f and f w f_w f w have isolated singularities, and by Corollary 2.1 the number μ k ( g ) \mu^k(g) μ k ( g ) , for g = f g=f g = f or g = f w g=f_w g = f w , depends only on k k k , μ ( g ) \mu(g) μ ( g ) and ν ( g ) \nu(g) ν ( g ) .
We claim that ν ( f w ) = ν ( f ) \nu(f_w)=\nu(f) ν ( f w ) = ν ( f ) . After exchanging the variables if necessary, assume that w 1 ≥ w 2 w_1 \ge w_2 w 1 ≥ w 2 . Since f w f_w f w has an isolated critical point, its gradient cannot vanish identically on the x x x -axis. Hence f w f_w f w contains a monomial x A y c x^A y^c x A y c with c ∈ { 0 , 1 } c \in\{0,1\} c ∈ { 0 , 1 } . Its weighted degree satisfies d = A w 1 + c w 2 > ( A + c − 1 ) w 1 d=Aw_1+cw_2>(A+c-1)w_1 d = A w 1 + c w 2 > ( A + c − 1 ) w 1 . Every monomial of total degree at most A + c − 1 A+c-1 A + c − 1 has weighted degree at most ( A + c − 1 ) w 1 < d (A+c-1)w_1<d ( A + c − 1 ) w 1 < d , so no such monomial occurs in either f w f_w f w or f ′ f' f ′ . Moreover, the monomial x A y c x^A y^c x A y c cannot be cancelled by f ′ f' f ′ , whose monomials have weighted degree strictly greater than d d d . Thus ν ( f w ) = ν ( f ) = A + c \nu(f_w)=\nu(f)=A+c ν ( f w ) = ν ( f ) = A + c .
Since μ ( f w ) = μ ( f ) \mu(f_w)=\mu(f) μ ( f w ) = μ ( f ) and ν ( f w ) = ν ( f ) \nu(f_w)=\nu(f) ν ( f w ) = ν ( f ) , we conclude that μ k ( f w ) = μ k ( f ) \mu^k(f_w)=\mu^k(f) μ k ( f w ) = μ k ( f ) for every integer k ≥ 0 k \ge0 k ≥ 0 . □
The following example illustrates Theorem 4.2 .
Example 4.3. Consider the function f ( x , y ) = x 3 + y 4 + x 2 y 3 f(x,y)=x^3+y^4+x^2y^3 f ( x , y ) = x 3 + y 4 + x 2 y 3 . Its semiquasihomogeneous principal part with respect to the weight vector w = ( 4 , 3 ) w=(4,3) w = ( 4 , 3 ) is f w ( x , y ) = x 3 + y 4 f_w(x,y)=x^3+y^4 f w ( x , y ) = x 3 + y 4 and x 2 y 3 x^2y^3 x 2 y 3 has weighted degree 17 > 12 17>12 17 > 12 . Here ν ( f ) = ν ( f w ) = 3 \nu(f)=\nu(f_w)=3 ν ( f ) = ν ( f w ) = 3 and, since J ( f w ) = ( x 2 , y 3 ) J(f_w)=(x^2,y^3) J ( f w ) = ( x 2 , y 3 ) , μ ( f ) = μ ( f w ) = dim C O 2 / ( x 2 , y 3 ) = 6 \mu(f)=\mu(f_w)=\dim_{\mathbb{C}}\mathcal{O}_2/(x^2,y^3)=6 μ ( f ) = μ ( f w ) = dim C O 2 / ( x 2 , y 3 ) = 6 . Using Corollary 2.1 , we get
μ k ( f ) = μ k ( f w ) = { 6 , k = 0 , k 2 + 5 k + 10 2 , k ≥ 1. \mu^k(f)=\mu^k(f_w)=
\begin{cases}
6, & k=0,\\
\frac{k^2+5k+10}{2}, & k\ge1.
\end{cases} μ k ( f ) = μ k ( f w ) = { 6 , 2 k 2 + 5 k + 10 , k = 0 , k ≥ 1.
Indeed, for k = 0 k=0 k = 0 we get μ 0 = 6 \mu^0=6 μ 0 = 6 . For k = 1 , 2 k=1,2 k = 1 , 2 (that is, k < 3 k<3 k < 3 ) Corollary 2.1 gives 6 + k ( k + 1 ) = ( k 2 + 5 k + 10 ) / 2 6+k(k+1)=(k^2+5k+10)/2 6 + k ( k + 1 ) = ( k 2 + 5 k + 10 ) /2 , and for k ≥ 3 k\ge3 k ≥ 3 it gives 6 + k ( k + 1 ) − ( k − 1 ) ( k − 2 ) / 2 = ( k 2 + 5 k + 10 ) / 2 6+k(k+1)-(k-1)(k-2)/2=(k^2+5k+10)/2 6 + k ( k + 1 ) − ( k − 1 ) ( k − 2 ) /2 = ( k 2 + 5 k + 10 ) /2 .
The k-th Milnor and Tjurina numbers of a pair of holomorphic function germs
Let f , g ∈ m ⊂ O 2 f,g\in\mathfrak{m}\subset\mathcal{O}_2 f , g ∈ m ⊂ O 2 and put P = g f x − f g x P=gf_x-fg_x P = g f x − f g x and Q = g f y − f g y Q=gf_y-fg_y Q = g f y − f g y . The Milnor number of the pair ( f , g ) (f,g) ( f , g ) is μ ( f , g ) = dim C O 2 / ( P , Q ) ∈ N ∪ { ∞ } \mu(f,g)=\dim_{\mathbb{C}}\mathcal{O}_2/(P,Q)\in\mathbb{N}\cup\{\infty\} μ ( f , g ) = dim C O 2 / ( P , Q ) ∈ N ∪ { ∞ } . When μ ( f , g ) < ∞ \mu(f,g)<\infty μ ( f , g ) < ∞ , we denote by F f / g \mathcal{F}_{f/g} F f / g the foliation defined by ω f / g = g d f − f d g \omega_{f/g}=g\,df-f\,dg ω f / g = g df − f d g .
For each integer k ≥ 0 k\ge0 k ≥ 0 , we define the k k k -th Milnor number of the pair ( f , g ) (f,g) ( f , g ) by
μ k ( f , g ) = dim C O 2 m k ( g f x − f g x , g f y − f g y ) ∈ N ∪ { ∞ } . \mu^k(f,g)=\dim_{\mathbb{C}}\frac{\mathcal{O}_2}{\mathfrak{m}^k(gf_x-fg_x,gf_y-fg_y)}\in\mathbb{N}\cup\{\infty\}. μ k ( f , g ) = dim C m k ( g f x − f g x , g f y − f g y ) O 2 ∈ N ∪ { ∞ } .
Note that μ 0 ( f , g ) = μ ( f , g ) \mu^0(f,g)=\mu(f,g) μ 0 ( f , g ) = μ ( f , g ) .
For f , g ∈ O 2 f,g\in\mathcal{O}_2 f , g ∈ O 2 , a point [ α : β ] ∈ P C 1 [\alpha:\beta]\in\mathbb{P}^1_{\mathbb{C}} [ α : β ] ∈ P C 1 and an integer k ≥ 0 k\geq0 k ≥ 0 , the k k k -th Tjurina number of the pair ( f , g ) (f,g) ( f , g ) with respect to C [ α : β ] : α f + β g = 0 C_{[\alpha:\beta]}:\alpha f+\beta g=0 C [ α : β ] : α f + β g = 0 is defined by
τ [ α : β ] k ( f , g ) = dim C O 2 m 2 k ( g f x − f g x , g f y − f g y ) + ( α f + β g ) . \tau^k_{[\alpha:\beta]}(f,g)=\dim_{\mathbb{C}}\frac{\mathcal{O}_2}{\mathfrak{m}_2^k(gf_x-fg_x,gf_y-fg_y)+(\alpha f+\beta g)}. τ [ α : β ] k ( f , g ) = dim C m 2 k ( g f x − f g x , g f y − f g y ) + ( α f + β g ) O 2 .
For integers k ≥ 0 k\geq0 k ≥ 0 and m 0 ≥ 0 m_0\geq0 m 0 ≥ 0 , put ε k ( m 0 ) = 0 \varepsilon_k(m_0)=0 ε k ( m 0 ) = 0 if 0 ≤ k < m 0 0\leq k<m_0 0 ≤ k < m 0 and ε k ( m 0 ) = ( k − m 0 + 1 ) ( k − m 0 ) 2 \varepsilon_k(m_0)=\dfrac{(k-m_0+1)(k-m_0)}{2} ε k ( m 0 ) = 2 ( k − m 0 + 1 ) ( k − m 0 ) if k ≥ m 0 k\geq m_0 k ≥ m 0 .
Theorem 5.1. Let f , g ∈ m ⊂ O 2 f,g\in\mathfrak{m}\subset\mathcal{O}_2 f , g ∈ m ⊂ O 2 be relatively prime reduced germs such that μ ( f , g ) < ∞ \mu(f,g)<\infty μ ( f , g ) < ∞ . Let m 0 = min { ν ( g f x − f g x ) , ν ( g f y − f g y ) } m_0=\min\{\nu(gf_x-fg_x),\nu(gf_y-fg_y)\} m 0 = min { ν ( g f x − f g x ) , ν ( g f y − f g y )} . Then, for every integer k ≥ 0 k\geq0 k ≥ 0 ,
μ k ( f , g ) = μ ( f , g ) + k ( k + 1 ) − ε k ( m 0 ) , \mu^k(f,g)=\mu(f,g)+k(k+1)-\varepsilon_k(m_0), μ k ( f , g ) = μ ( f , g ) + k ( k + 1 ) − ε k ( m 0 ) , that is,
μ k ( f , g ) = { μ ( f , g ) + k ( k + 1 ) , if 0 ≤ k < m 0 , μ ( f , g ) + k ( k + 1 ) − ( k − m 0 + 1 ) ( k − m 0 ) 2 , if k ≥ m 0 . \mu^k(f,g)=
\begin{cases}
\mu(f,g)+k(k+1), & \text{if }0\leq k<m_0,\\
\mu(f,g)+k(k+1)-\dfrac{(k-m_0+1)(k-m_0)}{2}, & \text{if }k\geq m_0.
\end{cases} μ k ( f , g ) = ⎩ ⎨ ⎧ μ ( f , g ) + k ( k + 1 ) , μ ( f , g ) + k ( k + 1 ) − 2 ( k − m 0 + 1 ) ( k − m 0 ) , if 0 ≤ k < m 0 , if k ≥ m 0 .
Proof. Since μ ( f , g ) < ∞ \mu(f,g)<\infty μ ( f , g ) < ∞ , the germs P P P and Q Q Q are relatively prime. Interchanging the names of P P P and Q Q Q if necessary (this does not change the ideal ( P , Q ) (P,Q) ( P , Q ) ), we may assume that ν ( P ) ≤ ν ( Q ) \nu(P)\leq\nu(Q) ν ( P ) ≤ ν ( Q ) , so that ν ( P ) = m 0 \nu(P)=m_0 ν ( P ) = m 0 .
Applying the formula from [21 Lemma 3.3] to the relatively prime germs P P P and Q Q Q , with ν ( P ) ≤ ν ( Q ) \nu(P)\leq\nu(Q) ν ( P ) ≤ ν ( Q ) , and J = m k J=\mathfrak{m}^k J = m k , we obtain
μ k ( F f / g ) = μ ( F f / g ) + dim O 2 m k + dim O 2 m k + ( P ) . (1) \mu^k(\mathcal{F}_{f/g})=\mu(\mathcal{F}_{f/g})+\dim\frac{\mathcal{O}_2}{\mathfrak{m}^k}+\dim\frac{\mathcal{O}_2}{\mathfrak{m}^k+(P)}. \tag*{(1)} μ k ( F f / g ) = μ ( F f / g ) + dim m k O 2 + dim m k + ( P ) O 2 . (1)
By definition, μ k ( F f / g ) = μ k ( f , g ) \mu^k(\mathcal{F}_{f/g})=\mu^k(f,g) μ k ( F f / g ) = μ k ( f , g ) and μ ( F f / g ) = μ ( f , g ) \mu(\mathcal{F}_{f/g})=\mu(f,g) μ ( F f / g ) = μ ( f , g ) . Moreover, dim O 2 / m k = k ( k + 1 ) / 2 \dim\mathcal{O}_2/\mathfrak{m}^k=k(k+1)/2 dim O 2 / m k = k ( k + 1 ) /2 , the number of monomials of degree less than k k k . Next, ( m k + ( P ) ) / m k ≅ O 2 / ( m k : P ) (\mathfrak{m}^k+(P))/\mathfrak{m}^k\cong\mathcal{O}_2/(\mathfrak{m}^k:P) ( m k + ( P )) / m k ≅ O 2 / ( m k : P ) . Since ν ( P a ) = m 0 + ν ( a ) \nu(Pa)=m_0+\nu(a) ν ( P a ) = m 0 + ν ( a ) for every a ∈ O 2 a\in\mathcal{O}_2 a ∈ O 2 , we have ( m k : P ) = m k − m 0 (\mathfrak{m}^k:P)=\mathfrak{m}^{k-m_0} ( m k : P ) = m k − m 0 if k ≥ m 0 k\geq m_0 k ≥ m 0 and ( m k : P ) = O 2 (\mathfrak{m}^k:P)=\mathcal{O}_2 ( m k : P ) = O 2 if k < m 0 k<m_0 k < m 0 . Hence dim ( m k + ( P ) ) / m k = ε k ( m 0 ) \dim(\mathfrak{m}^k+(P))/\mathfrak{m}^k=\varepsilon_k(m_0) dim ( m k + ( P )) / m k = ε k ( m 0 ) and
dim O 2 m k + ( P ) = k ( k + 1 ) 2 − ε k ( m 0 ) = { k ( k + 1 ) 2 , if 0 ≤ k < m 0 , k ( k + 1 ) 2 − ( k − m 0 + 1 ) ( k − m 0 ) 2 , if k ≥ m 0 . \dim\frac{\mathcal{O}_2}{\mathfrak{m}^k+(P)}
=\frac{k(k+1)}{2}-\varepsilon_k(m_0)
=
\begin{cases}
\dfrac{k(k+1)}{2}, & \text{if }0\leq k<m_0,\\
\dfrac{k(k+1)}{2}-\dfrac{(k-m_0+1)(k-m_0)}{2}, & \text{if }k\geq m_0.
\end{cases} dim m k + ( P ) O 2 = 2 k ( k + 1 ) − ε k ( m 0 ) = ⎩ ⎨ ⎧ 2 k ( k + 1 ) , 2 k ( k + 1 ) − 2 ( k − m 0 + 1 ) ( k − m 0 ) , if 0 ≤ k < m 0 , if k ≥ m 0 .
Substituting these expressions into (1) completes the proof of the theorem. □
Theorem 5.1 shows that μ k ( f , g ) \mu^k(f,g) μ k ( f , g ) depends only on μ ( f , g ) \mu(f,g) μ ( f , g ) and m 0 m_0 m 0 . Moreover, for f , g ∈ O 2 f,g\in\mathcal{O}_2 f , g ∈ O 2 and every integer k ≥ 0 k\geq0 k ≥ 0 , μ k ( f , g ) \mu^k(f,g) μ k ( f , g ) is invariant under biholomorphic changes of the source coordinates and invertible linear changes of the generators. Indeed, if
( f 1 g 1 ) = A ( f ∘ Φ g ∘ Φ ) , A ∈ GL 2 ( C ) , \begin{pmatrix}
f_1\\
g_1
\end{pmatrix}
=
A
\begin{pmatrix}
f\circ\Phi\\
g\circ\Phi
\end{pmatrix},
\qquad A\in\operatorname{GL}_2(\mathbb{C}), ( f 1 g 1 ) = A ( f ∘ Φ g ∘ Φ ) , A ∈ GL 2 ( C ) ,
where Φ \Phi Φ is a biholomorphic germ, then
g 1 d f 1 − f 1 d g 1 = det ( A ) Φ ∗ ( g d f − f d g ) . g_{1}\,df_{1}-f_{1}\,dg_{1}=\det(A)\,\Phi^{*}(g\,df-f\,dg). g 1 d f 1 − f 1 d g 1 = det ( A ) Φ ∗ ( g df − f d g ) .
The coefficients P 1 , Q 1 P_{1},Q_{1} P 1 , Q 1 of g 1 d f 1 − f 1 d g 1 g_{1}\,df_{1}-f_{1}\,dg_{1} g 1 d f 1 − f 1 d g 1 are det ( A ) \det(A) det ( A ) times the coefficients of Φ ∗ ( P d x + Q d y ) \Phi^{*}(P\,dx+Q\,dy) Φ ∗ ( P d x + Q d y ) , which are obtained from P ∘ Φ P\circ\Phi P ∘ Φ and Q ∘ Φ Q\circ\Phi Q ∘ Φ by multiplying by the Jacobian matrix of Φ \Phi Φ , invertible over O 2 \mathcal{O}_{2} O 2 . Hence ( P 1 , Q 1 ) = ( P ∘ Φ , Q ∘ Φ ) = Φ ∗ ( P , Q ) (P_{1},Q_{1})=(P\circ\Phi,Q\circ\Phi)=\Phi^{*}(P,Q) ( P 1 , Q 1 ) = ( P ∘ Φ , Q ∘ Φ ) = Φ ∗ ( P , Q ) . As Φ ∗ \Phi^{*} Φ ∗ is a ring automorphism of O 2 \mathcal{O}_{2} O 2 with Φ ∗ ( m ) = m \Phi^{*}(\mathfrak{m})=\mathfrak{m} Φ ∗ ( m ) = m , we obtain μ k ( f 1 , g 1 ) = μ k ( f , g ) \mu^{k}(f_{1},g_{1})=\mu^{k}(f,g) μ k ( f 1 , g 1 ) = μ k ( f , g ) for every integer k ≥ 0 k\geq0 k ≥ 0 .
Using the formula for the k k k -th Tjurina numbers established in [21 Theorem 8.2] , we relate τ [ α : β ] k ( f , g ) \tau^{k}_{[\alpha:\beta]}(f,g) τ [ α : β ] k ( f , g ) to the k k k -th Tjurina number of the curve C [ α : β ] C_{[\alpha:\beta]} C [ α : β ] .
Proposition 5.2. Let f , g ∈ m ⊂ O 2 f,g\in\mathfrak{m}\subset\mathcal{O}_{2} f , g ∈ m ⊂ O 2 be relatively prime reduced germs such that μ ( f , g ) < ∞ \mu(f,g)<\infty μ ( f , g ) < ∞ . Let [ α : β ] ∈ P C 1 [\alpha:\beta]\in\mathbb{P}^{1}_{\mathbb{C}} [ α : β ] ∈ P C 1 and let C [ α : β ] : α f + β g = 0 C_{[\alpha:\beta]}:\alpha f+\beta g=0 C [ α : β ] : α f + β g = 0 be the corresponding member of the pencil. Then C [ α : β ] C_{[\alpha:\beta]} C [ α : β ] is reduced and, for every integer k ≥ 0 k\geq0 k ≥ 0 , τ [ α : β ] k ( f , g ) = τ k ( C [ α : β ] ) + I 0 ( f , g ) \tau^{k}_{[\alpha:\beta]}(f,g)=\tau^{k}(C_{[\alpha:\beta]})+I_{0}(f,g) τ [ α : β ] k ( f , g ) = τ k ( C [ α : β ] ) + I 0 ( f , g ) .
Proof. Since μ ( f , g ) < ∞ \mu(f,g)<\infty μ ( f , g ) < ∞ , we have that F f / g \mathcal{F}_{f/g} F f / g is a foliation with an isolated singularity at the origin 0 ∈ C 2 0\in\mathbb{C}^{2} 0 ∈ C 2 . Put P = g f x − f g x P=g f_{x}-f g_{x} P = g f x − f g x and Q = g f y − f g y Q=g f_{y}-f g_{y} Q = g f y − f g y . Then τ [ α : β ] k ( f , g ) = τ k ( F f / g , C [ α : β ] ) \tau^{k}_{[\alpha:\beta]}(f,g)=\tau^{k}(\mathcal{F}_{f/g},C_{[\alpha:\beta]}) τ [ α : β ] k ( f , g ) = τ k ( F f / g , C [ α : β ] ) is the definition of the k k k -th Tjurina number of F f / g \mathcal{F}_{f/g} F f / g along C [ α : β ] C_{[\alpha:\beta]} C [ α : β ] . Moreover, C [ α : β ] C_{[\alpha:\beta]} C [ α : β ] is F f / g \mathcal{F}_{f/g} F f / g -invariant, since ω f / g ∧ d ( α f + β g ) = ( α f + β g ) d f ∧ d g \omega_{f/g}\wedge d(\alpha f+\beta g)=(\alpha f+\beta g)\,df\wedge dg ω f / g ∧ d ( α f + β g ) = ( α f + β g ) df ∧ d g is a multiple of α f + β g \alpha f+\beta g α f + β g . Since μ ( f , g ) < ∞ \mu(f,g)<\infty μ ( f , g ) < ∞ , the curve C [ α : β ] C_{[\alpha:\beta]} C [ α : β ] is reduced by [6 Lemma 4.1] , which proves the first assertion. Then, it follows from [21 Theorem 8.2] applied to F f / g \mathcal{F}_{f/g} F f / g and C [ α : β ] : α f + β g = 0 C_{[\alpha:\beta]}:\alpha f+\beta g=0 C [ α : β ] : α f + β g = 0 that
τ [ α : β ] k ( f , g ) = τ k ( F f / g , C [ α : β ] ) = τ k ( C [ α : β ] ) + GSV ( F f / g , C [ α : β ] ) , \tau^{k}_{[\alpha:\beta]}(f,g)=\tau^{k}(\mathcal{F}_{f/g},C_{[\alpha:\beta]})=\tau^{k}(C_{[\alpha:\beta]})+\operatorname{GSV}(\mathcal{F}_{f/g},C_{[\alpha:\beta]}), τ [ α : β ] k ( f , g ) = τ k ( F f / g , C [ α : β ] ) = τ k ( C [ α : β ] ) + GSV ( F f / g , C [ α : β ] ) ,
where GSV ( F f / g , C [ α : β ] ) \operatorname{GSV}(\mathcal{F}_{f/g},C_{[\alpha:\beta]}) GSV ( F f / g , C [ α : β ] ) denotes the GSV-index of F f / g \mathcal{F}_{f/g} F f / g with respect to C [ α : β ] C_{[\alpha:\beta]} C [ α : β ] (see [9 ] ). The proof ends by the equality GSV ( F f / g , C [ α : β ] ) = I 0 ( f , g ) \operatorname{GSV}(\mathcal{F}_{f/g},C_{[\alpha:\beta]})=I_{0}(f,g) GSV ( F f / g , C [ α : β ] ) = I 0 ( f , g ) , see [6 Proposition 4.8] . □
In particular, under the hypotheses of Proposition 5.2, the difference τ [ α : β ] k ( f , g ) − τ k ( C [ α : β ] ) = I 0 ( f , g ) \tau^{k}_{[\alpha:\beta]}(f,g)-\tau^{k}(C_{[\alpha:\beta]})=I_{0}(f,g) τ [ α : β ] k ( f , g ) − τ k ( C [ α : β ] ) = I 0 ( f , g ) is independent of both k k k and the chosen reduced member of the pencil.
For f , g ∈ O 2 f,g\in\mathcal{O}_{2} f , g ∈ O 2 such that μ ( f , g ) < ∞ \mu(f,g)<\infty μ ( f , g ) < ∞ , in [24 (2.1) Lemma] , Suwa studied the unfoldings of germs of meromorphic functions and introduced the invariant
η ( f , g ) = dim C ( f , g ) ( f x g − g x f , f y g − g y f ) . (2) \eta(f,g)=\dim_{\mathbb{C}}\frac{(f,g)}{(f_{x}g-g_{x}f,f_{y}g-g_{y}f)}. \tag*{(2)} η ( f , g ) = dim C ( f x g − g x f , f y g − g y f ) ( f , g ) . (2)
Lemma 5.3. Let f , g ∈ O 2 f,g\in\mathcal{O}_{2} f , g ∈ O 2 be such that μ ( f , g ) < ∞ \mu(f,g)<\infty μ ( f , g ) < ∞ . Then η ( f , g ) \eta(f,g) η ( f , g ) and I 0 ( f , g ) I_{0}(f,g) I 0 ( f , g ) are finite and
μ ( f , g ) = η ( f , g ) + I 0 ( f , g ) . (3) \mu(f,g)=\eta(f,g)+I_{0}(f,g). \tag*{(3)} μ ( f , g ) = η ( f , g ) + I 0 ( f , g ) . (3) In particular, I 0 ( f , g ) = μ ( f , g ) − η ( f , g ) I_{0}(f,g)=\mu(f,g)-\eta(f,g) I 0 ( f , g ) = μ ( f , g ) − η ( f , g ) .
Proof. Put P = g f x − f g x P=g f_{x}-f g_{x} P = g f x − f g x and Q = g f y − f g y Q=g f_{y}-f g_{y} Q = g f y − f g y . Since ( P , Q ) ⊆ ( f , g ) (P,Q)\subseteq(f,g) ( P , Q ) ⊆ ( f , g ) , there is an exact sequence of vector spaces
0 ⟶ ( f , g ) ( P , Q ) ⟶ O 2 ( P , Q ) ⟶ O 2 ( f , g ) ⟶ 0. 0\longrightarrow\frac{(f,g)}{(P,Q)}\longrightarrow\frac{\mathcal{O}_{2}}{(P,Q)}\longrightarrow\frac{\mathcal{O}_{2}}{(f,g)}\longrightarrow0. 0 ⟶ ( P , Q ) ( f , g ) ⟶ ( P , Q ) O 2 ⟶ ( f , g ) O 2 ⟶ 0.
Since μ ( f , g ) = dim O 2 / ( P , Q ) < ∞ \mu(f,g)=\dim\mathcal{O}_2/(P,Q)<\infty μ ( f , g ) = dim O 2 / ( P , Q ) < ∞ , the other two spaces have finite dimension, which are η ( f , g ) \eta(f,g) η ( f , g ) and I 0 ( f , g ) I_0(f,g) I 0 ( f , g ) respectively, and μ ( f , g ) = η ( f , g ) + I 0 ( f , g ) \mu(f,g)=\eta(f,g)+I_0(f,g) μ ( f , g ) = η ( f , g ) + I 0 ( f , g ) .
The following theorem relates μ k ( f , g ) \mu^k(f,g) μ k ( f , g ) and τ [ α : β ] k ( f , g ) \tau^k_{[\alpha:\beta]}(f,g) τ [ α : β ] k ( f , g ) .
Theorem 5.4. Let f , g ∈ m ⊂ O 2 f,g\in\mathfrak{m}\subset\mathcal{O}_2 f , g ∈ m ⊂ O 2 be relatively prime reduced germs such that μ ( f , g ) < ∞ \mu(f,g)<\infty μ ( f , g ) < ∞ . Let [ α : β ] ∈ P C 1 [\alpha:\beta]\in\mathbb{P}^1_{\mathbb{C}} [ α : β ] ∈ P C 1 and let C [ α : β ] : α f + β g = 0 C_{[\alpha:\beta]}:\alpha f+\beta g=0 C [ α : β ] : α f + β g = 0 be the corresponding member of the pencil, and let m 0 = min { ν ( g f x − f g x ) , ν ( g f y − f g y ) } m_0=\min\{\nu(gf_x-fg_x),\nu(gf_y-fg_y)\} m 0 = min { ν ( g f x − f g x ) , ν ( g f y − f g y )} . Then, for every integer k ≥ 0 k\ge0 k ≥ 0 ,
μ k ( f , g ) − τ [ α : β ] k ( f , g ) = η ( f , g ) + k ( k + 1 ) − τ k ( C [ α : β ] ) − ε k ( m 0 ) . \mu^k(f,g)-\tau^k_{[\alpha:\beta]}(f,g)=\eta(f,g)+k(k+1)-\tau^k(C_{[\alpha:\beta]})-\varepsilon_k(m_0). μ k ( f , g ) − τ [ α : β ] k ( f , g ) = η ( f , g ) + k ( k + 1 ) − τ k ( C [ α : β ] ) − ε k ( m 0 ) .
Proof. By Proposition 5.2 and Lemma 5.3 ,
τ [ α : β ] k ( f , g ) = τ k ( C [ α : β ] ) + I 0 ( f , g ) = τ k ( C [ α : β ] ) + μ ( f , g ) − η ( f , g ) . \tau^k_{[\alpha:\beta]}(f,g)=\tau^k(C_{[\alpha:\beta]})+I_0(f,g)=\tau^k(C_{[\alpha:\beta]})+\mu(f,g)-\eta(f,g). τ [ α : β ] k ( f , g ) = τ k ( C [ α : β ] ) + I 0 ( f , g ) = τ k ( C [ α : β ] ) + μ ( f , g ) − η ( f , g ) .
By Theorem 5.1 , μ k ( f , g ) = μ ( f , g ) + k ( k + 1 ) − ε k ( m 0 ) \mu^k(f,g)=\mu(f,g)+k(k+1)-\varepsilon_k(m_0) μ k ( f , g ) = μ ( f , g ) + k ( k + 1 ) − ε k ( m 0 ) . Subtracting, μ ( f , g ) \mu(f,g) μ ( f , g ) cancels and we obtain the formula.
The following example illustrates Theorem 5.4 .
Example 5.5. Let f ( x , y ) = x 2 + y 3 f(x,y)=x^2+y^3 f ( x , y ) = x 2 + y 3 and g ( x , y ) = x y g(x,y)=xy g ( x , y ) = x y in O 2 \mathcal{O}_2 O 2 . Consider the fiber
C [ α : β ] : α ( x 2 + y 3 ) + β x y = 0 , [ α : β ] ∈ P C 1 . C_{[\alpha:\beta]}:\alpha(x^2+y^3)+\beta xy=0,\qquad[\alpha:\beta]\in\mathbb{P}^1_{\mathbb{C}}. C [ α : β ] : α ( x 2 + y 3 ) + β x y = 0 , [ α : β ] ∈ P C 1 . The curve C [ 1 : 0 ] C_{[1:0]} C [ 1 : 0 ] is the cusp, and for [ α : β ] ≠ [ 1 : 0 ] [\alpha:\beta]\ne[1:0] [ α : β ] = [ 1 : 0 ] the curve C [ α : β ] C_{[\alpha:\beta]} C [ α : β ] is a node. In particular, all of them are reduced. We have τ k ( C [ 1 : 0 ] ) = 2 k + 2 \tau^k(C_{[1:0]})=2k+2 τ k ( C [ 1 : 0 ] ) = 2 k + 2 , and for [ α : β ] ≠ [ 1 : 0 ] [\alpha:\beta]\ne[1:0] [ α : β ] = [ 1 : 0 ] , τ k ( C [ α : β ] ) = 2 k + 1 \tau^k(C_{[\alpha:\beta]})=2k+1 τ k ( C [ α : β ] ) = 2 k + 1 . Moreover, P = x 2 y − y 4 P=x^2y-y^4 P = x 2 y − y 4 and Q = 2 x y 3 − x 3 Q=2xy^3-x^3 Q = 2 x y 3 − x 3 both have order 3 3 3 , and
m 0 = 3 , η ( f , g ) = 7 , μ ( f , g ) = 12 , I 0 ( f , g ) = 5. m_0=3,\qquad\eta(f,g)=7,\qquad\mu(f,g)=12,\qquad I_0(f,g)=5. m 0 = 3 , η ( f , g ) = 7 , μ ( f , g ) = 12 , I 0 ( f , g ) = 5. By Theorem 5.1 (and checked with Singular),
μ k ( f , g ) = { k 2 + k + 12 , 0 ≤ k < 3 , k 2 + 7 k + 18 2 , k ≥ 3. \mu^k(f,g)=
\begin{cases}
k^2+k+12, & 0\le k<3,\\
\dfrac{k^2+7k+18}{2}, & k\ge3.
\end{cases} μ k ( f , g ) = ⎩ ⎨ ⎧ k 2 + k + 12 , 2 k 2 + 7 k + 18 , 0 ≤ k < 3 , k ≥ 3. By Proposition 5.2 , τ [ α : β ] k ( f , g ) = 2 k + 6 \tau^k_{[\alpha:\beta]}(f,g)=2k+6 τ [ α : β ] k ( f , g ) = 2 k + 6 for [ α : β ] ≠ [ 1 : 0 ] [\alpha:\beta]\ne[1:0] [ α : β ] = [ 1 : 0 ] , and τ [ 1 : 0 ] k ( f , g ) = 2 k + 7 \tau^k_{[1:0]}(f,g)=2k+7 τ [ 1 : 0 ] k ( f , g ) = 2 k + 7 . Therefore
μ k ( f , g ) − τ [ α : β ] k ( f , g ) = { k 2 − k + 6 , 0 ≤ k < 3 , k 2 + 3 k + 6 2 , k ≥ 3. \mu^k(f,g)-\tau^k_{[\alpha:\beta]}(f,g)=
\begin{cases}
k^2-k+6, & 0\le k<3,\\
\dfrac{k^2+3k+6}{2}, & k\ge3.
\end{cases} μ k ( f , g ) − τ [ α : β ] k ( f , g ) = ⎩ ⎨ ⎧ k 2 − k + 6 , 2 k 2 + 3 k + 6 , 0 ≤ k < 3 , k ≥ 3. and
μ k ( f , g ) − τ [ 1 : 0 ] k ( f , g ) = { k 2 − k + 5 , 0 ≤ k < 3 , k 2 + 3 k + 4 2 , k ≥ 3. \mu^k(f,g)-\tau^k_{[1:0]}(f,g)=
\begin{cases}
k^2-k+5, & 0\le k<3,\\
\dfrac{k^2+3k+4}{2}, & k\ge3.
\end{cases} μ k ( f , g ) − τ [ 1 : 0 ] k ( f , g ) = ⎩ ⎨ ⎧ k 2 − k + 5 , 2 k 2 + 3 k + 4 , 0 ≤ k < 3 , k ≥ 3. respectively.
Bifurcation formula for the pair
In this section we use Theorem 5.1 and Corollary 2.2 to compare the k k k -th Milnor number μ k ( f , g ) \mu^{k}(f,g) μ k ( f , g ) of the pair with the k k k -th Milnor number μ k ( f g ) \mu^{k}(fg) μ k ( f g ) of the product, in the spirit of the bifurcation formula of [6 ] . We first compute the algebraic multiplicity of the foliation F f / g \mathcal{F}_{f/g} F f / g in terms of the bifurcation fibers of the pencil.
Let f , g ∈ m ⊂ O 2 f,g \in\mathfrak{m} \subset\mathcal{O}_{2} f , g ∈ m ⊂ O 2 be relatively prime germs such that μ ( f , g ) < ∞ \mu(f,g)<\infty μ ( f , g ) < ∞ . Let C [ α : β ] = { α f + β g = 0 } C_{[\alpha:\beta]}=\{\alpha f+\beta g=0\} C [ α : β ] = { α f + β g = 0 } denote the member of the pencil P f , g \mathcal{P}_{f,g} P f , g defined by [ α : β ] ∈ P C 1 [\alpha:\beta]\in\mathbb{P}^{1}_{\mathbb{C}} [ α : β ] ∈ P C 1 . By [6 Lemma 4.1] , every C [ α : β ] C_{[\alpha:\beta]} C [ α : β ] is reduced, so μ ( C [ α : β ] ) < ∞ \mu(C_{[\alpha:\beta]})<\infty μ ( C [ α : β ] ) < ∞ . The generic Milnor number and the generic multiplicity of the pencil are
μ g e n = min [ α : β ] ∈ P C 1 μ ( C [ α : β ] ) and ν g e n = min [ α : β ] ∈ P C 1 ν ( α f + β g ) . \mu_{\mathrm{gen}}=\min_{[\alpha:\beta]\in\mathbb{P}^{1}_{\mathbb{C}}}\mu(C_{[\alpha:\beta]})\quad\text{and}\quad\nu_{\mathrm{gen}}=\min_{[\alpha:\beta]\in\mathbb{P}^{1}_{\mathbb{C}}}\nu(\alpha f+\beta g). μ gen = [ α : β ] ∈ P C 1 min μ ( C [ α : β ] ) and ν gen = [ α : β ] ∈ P C 1 min ν ( α f + β g ) .
The bifurcation set is the finite set (its finiteness is proved in [6 ] )
B ( f , g ) = { [ α : β ] ∈ P C 1 : μ ( C [ α : β ] ) > μ g e n } , \mathcal{B}(f,g)=\{[\alpha:\beta]\in\mathbb{P}^{1}_{\mathbb{C}}:\mu(C_{[\alpha:\beta]})>\mu_{\mathrm{gen}}\}, B ( f , g ) = {[ α : β ] ∈ P C 1 : μ ( C [ α : β ] ) > μ gen } ,
and we put
B ∗ ( f , g ) = B ( f , g ) ∖ { [ 1 : 0 ] , [ 0 : 1 ] } . \mathcal{B}^{*}(f,g)=\mathcal{B}(f,g)\setminus\{[1:0],[0:1]\}. B ∗ ( f , g ) = B ( f , g ) ∖ {[ 1 : 0 ] , [ 0 : 1 ]} .
Note that [ 1 : 0 ] [1:0] [ 1 : 0 ] and [ 0 : 1 ] [0:1] [ 0 : 1 ] correspond to f = 0 f=0 f = 0 and g = 0 g=0 g = 0 . For [ α : β ] ∈ B ∗ ( f , g ) [\alpha:\beta]\in\mathcal{B}^{*}(f,g) [ α : β ] ∈ B ∗ ( f , g ) we call C [ α : β ] C_{[\alpha:\beta]} C [ α : β ] a bifurcation fiber of P f , g \mathcal{P}_{f,g} P f , g other than f f f and g g g .
The following proposition was established in [8 Corollary 5.3] ; here we provide a different proof that does not use the bifurcation formula given in [6 ] .
Proposition 6.1. Let f , g ∈ m ⊂ O 2 f,g\in\mathfrak{m}\subset\mathcal{O}_{2} f , g ∈ m ⊂ O 2 be relatively prime germs such that μ ( f , g ) < ∞ \mu(f,g)<\infty μ ( f , g ) < ∞ , and let h 1 , … , h r h_{1},\ldots,h_{r} h 1 , … , h r be the distinct bifurcation fibers of P f , g \mathcal{P}_{f,g} P f , g other than f f f and g g g . Then
ν ( F f / g ) = ν ( f ) + ν ( g ) − 1 + ∑ j = 1 r ( ν ( h j ) − ν g e n ) . \nu(\mathcal{F}_{f/g})=\nu(f)+\nu(g)-1+\sum_{j=1}^{r}\bigl(\nu(h_{j})-\nu_{\mathrm{gen}}\bigr). ν ( F f / g ) = ν ( f ) + ν ( g ) − 1 + j = 1 ∑ r ( ν ( h j ) − ν gen ) . Moreover, at most one summand is nonzero, and this happens if and only if ν ( f ) = ν ( g ) \nu(f)=\nu(g) ν ( f ) = ν ( g ) and the initial forms of f f f and g g g are proportional.
Proof. Put M = ν ( f ) + ν ( g ) M=\nu(f)+\nu(g) M = ν ( f ) + ν ( g ) , P = g f x − f g x P=gf_{x}-fg_{x} P = g f x − f g x and Q = g f y − f g y Q=gf_{y}-fg_{y} Q = g f y − f g y . As in the proof of Theorem 6.2 below, ν ( P ) , ν ( Q ) ≥ M − 1 \nu(P),\nu(Q)\geq M-1 ν ( P ) , ν ( Q ) ≥ M − 1 . Let P M − 1 P_{M-1} P M − 1 and Q M − 1 Q_{M-1} Q M − 1 be the homogeneous parts of degree M − 1 M-1 M − 1 of P P P and Q Q Q . By Euler’s identity,
x P M − 1 + y Q M − 1 = ( ν ( f ) − ν ( g ) ) in ( f ) in ( g ) . xP_{M-1}+yQ_{M-1}=(\nu(f)-\nu(g))\operatorname{in}(f)\operatorname{in}(g). x P M − 1 + y Q M − 1 = ( ν ( f ) − ν ( g )) in ( f ) in ( g ) .
If ν ( f ) ≠ ν ( g ) \nu(f)\ne\nu(g) ν ( f ) = ν ( g ) , then ( P M − 1 , Q M − 1 ) ≠ ( 0 , 0 ) (P_{M-1},Q_{M-1})\ne(0,0) ( P M − 1 , Q M − 1 ) = ( 0 , 0 ) , so ν ( F f / g ) = M − 1 \nu(\mathcal{F}_{f/g})=M-1 ν ( F f / g ) = M − 1 . If ν ( f ) = ν ( g ) \nu(f)=\nu(g) ν ( f ) = ν ( g ) and the initial forms of f f f and g g g are not proportional, then P M − 1 = Q M − 1 = 0 P_{M-1}=Q_{M-1}=0 P M − 1 = Q M − 1 = 0 would give in ( g ) d in ( f ) − in ( f ) d in ( g ) = 0 \operatorname{in}(g)\,d\operatorname{in}(f)-\operatorname{in}(f)\,d\operatorname{in}(g)=0 in ( g ) d in ( f ) − in ( f ) d in ( g ) = 0 , that is, d ( in ( f ) / in ( g ) ) = 0 d(\operatorname{in}(f)/\operatorname{in}(g))=0 d ( in ( f ) / in ( g )) = 0 , a contradiction; hence again ν ( F f / g ) = M − 1 \nu(\mathcal{F}_{f/g})=M-1 ν ( F f / g ) = M − 1 . In both cases in ( α f + β g ) ≠ 0 \operatorname{in}(\alpha f+\beta g)\ne0 in ( α f + β g ) = 0 has degree min { ν ( f ) , ν ( g ) } \min\{\nu(f),\nu(g)\} min { ν ( f ) , ν ( g )} for every [ α : β ] ∉ { [ 1 : 0 ] , [ 0 : 1 ] } [\alpha:\beta]\notin\{[1:0],[0:1]\} [ α : β ] ∈ / {[ 1 : 0 ] , [ 0 : 1 ]} , so every member of P f , g \mathcal{P}_{f,g} P f , g other than f f f and g g g has multiplicity ν g e n \nu_{\mathrm{gen}} ν gen , and the sum vanishes.
Assume now that ν ( f ) = ν ( g ) \nu(f)=\nu(g) ν ( f ) = ν ( g ) and in ( f ) = c in ( g ) \operatorname{in}(f)=c\operatorname{in}(g) in ( f ) = c in ( g ) with c ∈ C ∗ c\in\mathbb{C}^{*} c ∈ C ∗ . Then ν g e n = ν ( f ) = ν ( g ) \nu_{\mathrm{gen}}=\nu(f)=\nu(g) ν gen = ν ( f ) = ν ( g ) , and h = f − c g h=f-cg h = f − c g satisfies ν ( h ) > ν g e n \nu(h)>\nu_{\mathrm{gen}} ν ( h ) > ν gen . It is the only member of P f , g \mathcal{P}_{f,g} P f , g with this property: two distinct such members would generate P f , g \mathcal{P}_{f,g} P f , g , and then every member would have multiplicity greater than ν g e n \nu_{\mathrm{gen}} ν gen . Since g d h − h d g = g d f − f d g g\,dh-h\,dg=g\,df-f\,dg g d h − h d g = g df − f d g and ν ( h ) ≠ ν ( g ) \nu(h)\ne\nu(g) ν ( h ) = ν ( g ) , the case ν ( f ) ≠ ν ( g ) \nu(f)\ne\nu(g) ν ( f ) = ν ( g ) applied to the pair ( h , g ) (h,g) ( h , g ) gives
ν ( F f / g ) = ν ( h ) + ν ( g ) − 1 = ν ( f ) + ν ( g ) − 1 + ( ν ( h ) − ν g e n ) . \nu(F_{f/g})=\nu(h)+\nu(g)-1=\nu(f)+\nu(g)-1+(\nu(h)-\nu_{\mathrm{gen}}). ν ( F f / g ) = ν ( h ) + ν ( g ) − 1 = ν ( f ) + ν ( g ) − 1 + ( ν ( h ) − ν gen ) .
It remains to show that h h h is a bifurcation fiber. Suppose that μ ( h ) = μ g e n \mu(h)=\mu_{\mathrm{gen}} μ ( h ) = μ gen . Since B ( f , g ) \mathcal{B}(f,g) B ( f , g ) is finite, the family h s = f − ( c + s ) g h_s=f-(c+s)g h s = f − ( c + s ) g , s ∈ ( C , 0 ) s\in(\mathbb{C},0) s ∈ ( C , 0 ) , consists of reduced curves with μ ( h s ) = μ g e n \mu(h_s)=\mu_{\mathrm{gen}} μ ( h s ) = μ gen for every s s s . By the Lê–Ramanujam theorem [15 ] , this μ \mu μ -constant family is topologically trivial, and since the multiplicity of a reduced plane curve germ is a topological invariant (see [11 ] ), all its members have the same multiplicity. This contradicts ν ( h s ) = ν g e n < ν ( h ) = ν ( h 0 ) \nu(h_s)=\nu_{\mathrm{gen}}<\nu(h)=\nu(h_0) ν ( h s ) = ν gen < ν ( h ) = ν ( h 0 ) for s ≠ 0 s\ne0 s = 0 . Hence μ ( h ) > μ g e n \mu(h)>\mu_{\mathrm{gen}} μ ( h ) > μ gen , so h ∈ { h 1 , … , h r } h\in\{h_1,\ldots,h_r\} h ∈ { h 1 , … , h r } , while the other h j h_j h j have multiplicity ν g e n \nu_{\mathrm{gen}} ν gen . □
Theorem 6.2. Let f , g ∈ m ⊂ O 2 f,g\in\mathfrak{m}\subset\mathcal{O}_2 f , g ∈ m ⊂ O 2 be relatively prime reduced germs such that μ ( f , g ) < ∞ \mu(f,g)<\infty μ ( f , g ) < ∞ . Let m = ν ( f ) m=\nu(f) m = ν ( f ) , ℓ = ν ( g ) \ell=\nu(g) ℓ = ν ( g ) , M = m + ℓ M=m+\ell M = m + ℓ and m 0 = min { ν ( g f x − f g x ) , ν ( g f y − f g y ) } m_0=\min\{\nu(gf_x-fg_x),\nu(gf_y-fg_y)\} m 0 = min { ν ( g f x − f g x ) , ν ( g f y − f g y )} , and put
Δ = ∑ [ α : β ] ∈ B ∗ ( f , g ) ( μ ( C [ α : β ] ) − μ g e n ) and δ = ∑ [ α : β ] ∈ B ∗ ( f , g ) ( ν ( α f + β g ) − ν g e n ) . \Delta=\sum_{[\alpha:\beta]\in\mathcal{B}^{*}(f,g)}\bigl(\mu(C_{[\alpha:\beta]})-\mu_{\mathrm{gen}}\bigr)
\quad\text{and}\quad
\delta=\sum_{[\alpha:\beta]\in\mathcal{B}^{*}(f,g)}\bigl(\nu(\alpha f+\beta g)-\nu_{\mathrm{gen}}\bigr). Δ = [ α : β ] ∈ B ∗ ( f , g ) ∑ ( μ ( C [ α : β ] ) − μ gen ) and δ = [ α : β ] ∈ B ∗ ( f , g ) ∑ ( ν ( α f + β g ) − ν gen ) . Then m 0 = M − 1 + δ ≥ M − 1 m_0=M-1+\delta\ge M-1 m 0 = M − 1 + δ ≥ M − 1 and, for every integer k ≥ 0 k\ge0 k ≥ 0 ,
μ k ( f , g ) = μ k ( f g ) + Δ + ε k ( M − 1 ) − ε k ( M − 1 + δ ) . \mu^{k}(f,g)=\mu^{k}(fg)+\Delta+\varepsilon_k(M-1)-\varepsilon_k(M-1+\delta). μ k ( f , g ) = μ k ( f g ) + Δ + ε k ( M − 1 ) − ε k ( M − 1 + δ ) . Proof. Put P = g f x − f g x P=gf_x-fg_x P = g f x − f g x and Q = g f y − f g y Q=gf_y-fg_y Q = g f y − f g y . Since ν ( f x ) , ν ( f y ) ≥ m − 1 \nu(f_x),\nu(f_y)\ge m-1 ν ( f x ) , ν ( f y ) ≥ m − 1 and ν ( g x ) , ν ( g y ) ≥ ℓ − 1 \nu(g_x),\nu(g_y)\ge\ell-1 ν ( g x ) , ν ( g y ) ≥ ℓ − 1 , we have ν ( g f x ) , ν ( f g x ) , ν ( g f y ) , ν ( f g y ) ≥ M − 1 \nu(gf_x),\nu(fg_x),\nu(gf_y),\nu(fg_y)\ge M-1 ν ( g f x ) , ν ( f g x ) , ν ( g f y ) , ν ( f g y ) ≥ M − 1 . Hence ν ( P ) ≥ M − 1 \nu(P)\ge M-1 ν ( P ) ≥ M − 1 and ν ( Q ) ≥ M − 1 \nu(Q)\ge M-1 ν ( Q ) ≥ M − 1 , that is, m 0 ≥ M − 1 m_0\ge M-1 m 0 ≥ M − 1 . Since m 0 = ν ( F f / g ) m_0=\nu(F_{f/g}) m 0 = ν ( F f / g ) , Proposition 6.1 gives m 0 = M − 1 + δ m_0=M-1+\delta m 0 = M − 1 + δ . By the bifurcation formula of [6 Theorem 4.6] , μ ( f , g ) = μ ( f g ) + Δ \mu(f,g)=\mu(fg)+\Delta μ ( f , g ) = μ ( f g ) + Δ . Since f f f and g g g are reduced and relatively prime, Corollary 2.2 gives μ k ( f g ) = μ ( f g ) + k ( k + 1 ) − ε k ( M − 1 ) \mu^{k}(fg)=\mu(fg)+k(k+1)-\varepsilon_k(M-1) μ k ( f g ) = μ ( f g ) + k ( k + 1 ) − ε k ( M − 1 ) . By Theorem 5.1, μ k ( f , g ) = μ ( f , g ) + k ( k + 1 ) − ε k ( m 0 ) \mu^{k}(f,g)=\mu(f,g)+k(k+1)-\varepsilon_k(m_0) μ k ( f , g ) = μ ( f , g ) + k ( k + 1 ) − ε k ( m 0 ) . Subtracting, and using μ ( f , g ) − μ ( f g ) = Δ \mu(f,g)-\mu(fg)=\Delta μ ( f , g ) − μ ( f g ) = Δ ,
μ k ( f , g ) − μ k ( f g ) = Δ + ε k ( M − 1 ) − ε k ( m 0 ) . \mu^{k}(f,g)-\mu^{k}(fg)=\Delta+\varepsilon_k(M-1)-\varepsilon_k(m_0). μ k ( f , g ) − μ k ( f g ) = Δ + ε k ( M − 1 ) − ε k ( m 0 ) .
Remark 6.3. By Proposition 6.1, δ > 0 \delta>0 δ > 0 if and only if m = ℓ m=\ell m = ℓ and the initial forms of f f f and g g g are proportional; in this case δ = ν ( h ) − ν g e n \delta=\nu(h)-\nu_{\mathrm{gen}} δ = ν ( h ) − ν gen , where h h h is the unique member of P f , g \mathcal{P}_{f,g} P f , g of multiplicity greater than ν g e n \nu_{\mathrm{gen}} ν gen . Substituting the definition of ε k \varepsilon_k ε k in the formula of Theorem 6.2 gives, according to the value of δ \delta δ :
If δ = 0 \delta=0 δ = 0 , that is, m 0 = M − 1 m_0=M-1 m 0 = M − 1 , then ε k ( M − 1 ) − ε k ( m 0 ) = 0 \varepsilon_k(M-1)-\varepsilon_k(m_0)=0 ε k ( M − 1 ) − ε k ( m 0 ) = 0 , and so
μ k ( f , g ) = μ k ( f g ) + Δ for every k ≥ 0. \mu^{k}(f,g)=\mu^{k}(fg)+\Delta\quad\text{for every }k\ge0. μ k ( f , g ) = μ k ( f g ) + Δ for every k ≥ 0.
(2) If δ = 1 \delta=1 δ = 1 , that is, m 0 = M m_0=M m 0 = M , then ε k ( M − 1 ) = ε k ( M ) = 0 \varepsilon_k(M-1)=\varepsilon_k(M)=0 ε k ( M − 1 ) = ε k ( M ) = 0 for k < M k<M k < M , and ε k ( M − 1 ) − ε k ( M ) = k − M + 1 \varepsilon_k(M-1)-\varepsilon_k(M)=k-M+1 ε k ( M − 1 ) − ε k ( M ) = k − M + 1 for k ≥ M k\ge M k ≥ M . Hence
μ k ( f , g ) = { μ k ( f g ) + Δ , if 0 ≤ k < M , μ k ( f g ) + Δ + ( k − M + 1 ) , if k ≥ M . \mu^k(f,g)=
\begin{cases}
\mu^k(fg)+\Delta, & \text{if }0\le k<M,\\
\mu^k(fg)+\Delta+(k-M+1), & \text{if }k\ge M.
\end{cases} μ k ( f , g ) = { μ k ( f g ) + Δ , μ k ( f g ) + Δ + ( k − M + 1 ) , if 0 ≤ k < M , if k ≥ M .
(3) If δ ≥ 2 \delta\ge2 δ ≥ 2 , that is, m 0 > M m_0>M m 0 > M , then ε k ( M − 1 ) = ε k ( m 0 ) = 0 \varepsilon_k(M-1)=\varepsilon_k(m_0)=0 ε k ( M − 1 ) = ε k ( m 0 ) = 0 for k < M k<M k < M ; for M ≤ k < m 0 M\le k<m_0 M ≤ k < m 0 , ε k ( m 0 ) = 0 \varepsilon_k(m_0)=0 ε k ( m 0 ) = 0 and ε k ( M − 1 ) = ( k − M + 2 ) ( k − M + 1 ) 2 \varepsilon_k(M-1)=\frac{(k-M+2)(k-M+1)}{2} ε k ( M − 1 ) = 2 ( k − M + 2 ) ( k − M + 1 ) ; for k ≥ m 0 k\ge m_0 k ≥ m 0 both are given by their second expression. Hence
μ k ( f , g ) = { μ k ( f g ) + Δ , if 0 ≤ k < M , μ k ( f g ) + Δ + ( k − M + 2 ) ( k − M + 1 ) 2 , if M ≤ k < m 0 , μ k ( f g ) + Δ + ( k − M + 2 ) ( k − M + 1 ) − ( k − m 0 + 1 ) ( k − m 0 ) 2 , if k ≥ m 0 . \mu^k(f,g)=
\begin{cases}
\mu^k(fg)+\Delta, & \text{if }0\le k<M,\\
\mu^k(fg)+\Delta+\frac{(k-M+2)(k-M+1)}{2}, & \text{if }M\le k<m_0,\\
\mu^k(fg)+\Delta+\frac{(k-M+2)(k-M+1)-(k-m_0+1)(k-m_0)}{2}, & \text{if }k\ge m_0.
\end{cases} μ k ( f , g ) = ⎩ ⎨ ⎧ μ k ( f g ) + Δ , μ k ( f g ) + Δ + 2 ( k − M + 2 ) ( k − M + 1 ) , μ k ( f g ) + Δ + 2 ( k − M + 2 ) ( k − M + 1 ) − ( k − m 0 + 1 ) ( k − m 0 ) , if 0 ≤ k < M , if M ≤ k < m 0 , if k ≥ m 0 .
The following example illustrates Proposition 6.1 and Theorem 6.2 .
Example 6.4. (1) Let f = x 2 + y 2 f=x^2+y^2 f = x 2 + y 2 and g = x 2 + 2 y 2 + x 3 + y 3 g=x^2+2y^2+x^3+y^3 g = x 2 + 2 y 2 + x 3 + y 3 , so m = ℓ = 2 m=\ell=2 m = ℓ = 2 . The member α f + β g \alpha f+\beta g α f + β g of the pencil has quadratic part ( α + β ) x 2 + ( α + 2 β ) y 2 (\alpha+\beta)x^2+(\alpha+2\beta)y^2 ( α + β ) x 2 + ( α + 2 β ) y 2 , so it is a node unless [ α : β ] ∈ { [ 1 : − 1 ] , [ 2 : − 1 ] } [\alpha:\beta]\in\{[1:-1],[2:-1]\} [ α : β ] ∈ {[ 1 : − 1 ] , [ 2 : − 1 ]} . The fibers of the pencil α f + β g \alpha f+\beta g α f + β g are nodes, except
h 1 = f − g = − ( y 2 + x 3 + y 3 ) , h 2 = f − 1 2 g = 1 2 ( x 2 − x 3 − y 3 ) , h_1=f-g=-(y^2+x^3+y^3),\qquad h_2=f-\frac{1}{2}g=\frac{1}{2}(x^2-x^3-y^3), h 1 = f − g = − ( y 2 + x 3 + y 3 ) , h 2 = f − 2 1 g = 2 1 ( x 2 − x 3 − y 3 ) , which are cusps. Hence B ∗ ( f , g ) = { [ 1 : − 1 ] , [ 2 : − 1 ] } B^*(f,g)=\{[1:-1],[2:-1]\} B ∗ ( f , g ) = {[ 1 : − 1 ] , [ 2 : − 1 ]} , μ ( h 1 ) = μ ( h 2 ) = 2 \mu(h_1)=\mu(h_2)=2 μ ( h 1 ) = μ ( h 2 ) = 2 and μ g e n = 1 \mu_{\mathrm{gen}}=1 μ gen = 1 , so Δ = ( 2 − 1 ) + ( 2 − 1 ) = 2 \Delta=(2-1)+(2-1)=2 Δ = ( 2 − 1 ) + ( 2 − 1 ) = 2 . Since ν ( h 1 ) = ν ( h 2 ) = 2 = ν g e n \nu(h_1)=\nu(h_2)=2=\nu_{\mathrm{gen}} ν ( h 1 ) = ν ( h 2 ) = 2 = ν gen , we have δ = 0 \delta=0 δ = 0 , and Proposition 6.1 gives m 0 = m + ℓ − 1 = 3 m_0=m+\ell-1=3 m 0 = m + ℓ − 1 = 3 . Indeed, the initial forms of P = g f x − f g x P=gf_x-fg_x P = g f x − f g x and Q = g f y − f g y Q=gf_y-fg_y Q = g f y − f g y are 2 x y 2 2xy^2 2 x y 2 and − 2 x 2 y -2x^2y − 2 x 2 y , respectively. This example shows that a bifurcation fiber may have the generic multiplicity. Since f f f and g g g are nodes whose tangent cones have no common line, μ ( f ) = μ ( g ) = 1 \mu(f)=\mu(g)=1 μ ( f ) = μ ( g ) = 1 and I 0 ( f , g ) = 4 I_0(f,g)=4 I 0 ( f , g ) = 4 , so μ ( f g ) = μ ( f ) + μ ( g ) + 2 I 0 ( f , g ) − 1 = 9 \mu(fg)=\mu(f)+\mu(g)+2I_0(f,g)-1=9 μ ( f g ) = μ ( f ) + μ ( g ) + 2 I 0 ( f , g ) − 1 = 9 . By Theorem 6.2 ,
μ k ( f , g ) = μ k ( f g ) + 2 for all k ≥ 0 , \mu^k(f,g)=\mu^k(fg)+2\qquad\text{for all }k\ge0, μ k ( f , g ) = μ k ( f g ) + 2 for all k ≥ 0 , where
μ k ( f g ) = 9 + k ( k + 1 ) − ε k ( 3 ) and μ k ( f , g ) = 11 + k ( k + 1 ) − ε k ( 3 ) , \mu^k(fg)=9+k(k+1)-\varepsilon_k(3)\qquad\text{and}\qquad\mu^k(f,g)=11+k(k+1)-\varepsilon_k(3), μ k ( f g ) = 9 + k ( k + 1 ) − ε k ( 3 ) and μ k ( f , g ) = 11 + k ( k + 1 ) − ε k ( 3 ) , with ε k ( 3 ) = 0 \varepsilon_k(3)=0 ε k ( 3 ) = 0 if 0 ≤ k < 3 0\le k<3 0 ≤ k < 3 and ε k ( 3 ) = ( k − 2 ) ( k − 3 ) 2 \varepsilon_k(3)=\frac{(k-2)(k-3)}{2} ε k ( 3 ) = 2 ( k − 2 ) ( k − 3 ) if k ≥ 3 k\ge3 k ≥ 3 . Since δ = 0 \delta=0 δ = 0 , the two formulas have the same correction term ε k ( 3 ) \varepsilon_k(3) ε k ( 3 ) , and their difference is the constant Δ = 2 \Delta=2 Δ = 2 .
(2) Let f = x y + x 3 f=xy+x^{3} f = x y + x 3 and g = x y + y 3 g=xy+y^{3} g = x y + y 3 , so m = ℓ = 2 m=\ell=2 m = ℓ = 2 . Here f g = x y ( y + x 2 ) ( x + y 2 ) fg=xy(y+x^{2})(x+y^{2}) f g = x y ( y + x 2 ) ( x + y 2 ) , and μ ( f g ) = μ ( f ) + μ ( g ) + 2 I 0 ( f , g ) − 1 = 13 \mu(fg)=\mu(f)+\mu(g)+2I_{0}(f,g)-1=13 μ ( f g ) = μ ( f ) + μ ( g ) + 2 I 0 ( f , g ) − 1 = 13 , since μ ( f ) = μ ( g ) = 1 \mu(f)=\mu(g)=1 μ ( f ) = μ ( g ) = 1 and I 0 ( f , g ) = 6 I_{0}(f,g)=6 I 0 ( f , g ) = 6 . The member α f + β g = ( α + β ) x y + α x 3 + β y 3 \alpha f+\beta g=(\alpha+\beta)xy+\alpha x^{3}+\beta y^{3} α f + β g = ( α + β ) x y + α x 3 + β y 3 is a node unless α + β = 0 \alpha+\beta=0 α + β = 0 . The fibers of the pencil are nodes, except h = f − g = x 3 − y 3 h=f-g=x^{3}-y^{3} h = f − g = x 3 − y 3 , three lines with μ ( h ) = 4 \mu(h)=4 μ ( h ) = 4 . Hence B ∗ ( f , g ) = { [ 1 : − 1 ] } \mathcal{B}^{*}(f,g)=\{[1:-1]\} B ∗ ( f , g ) = {[ 1 : − 1 ]} , μ g e n = 1 \mu_{\mathrm{gen}}=1 μ gen = 1 and Δ = 4 − 1 = 3 \Delta=4-1=3 Δ = 4 − 1 = 3 . Since ν ( h ) = 3 \nu(h)=3 ν ( h ) = 3 and ν g e n = 2 \nu_{\mathrm{gen}}=2 ν gen = 2 , we have δ = 1 \delta=1 δ = 1 , and Proposition 6.1 gives m 0 = m + ℓ = 4 m_{0}=m+\ell=4 m 0 = m + ℓ = 4 . Indeed, the initial forms of P P P and Q Q Q are 2 x 3 y + y 4 2x^{3}y+y^{4} 2 x 3 y + y 4 and − 2 x y 3 − x 4 -2xy^{3}-x^{4} − 2 x y 3 − x 4 , respectively. By Theorem 6.2 ,
μ k ( f , g ) = { μ k ( f g ) + 3 , 0 ≤ k < 4 , μ k ( f g ) + k , k ≥ 4. \mu^{k}(f,g)=
\begin{cases}
\mu^{k}(fg)+3, & 0\leq k<4,\\
\mu^{k}(fg)+k, & k\geq4.
\end{cases} μ k ( f , g ) = { μ k ( f g ) + 3 , μ k ( f g ) + k , 0 ≤ k < 4 , k ≥ 4.
where
μ k ( f g ) = { 13 + k ( k + 1 ) , 0 ≤ k < 4 , k 2 + 7 k + 20 2 , k ≥ 4 , and μ k ( f , g ) = { 16 + k ( k + 1 ) , 0 ≤ k < 4 , k 2 + 9 k + 20 2 , k ≥ 4. \mu^{k}(fg)=
\begin{cases}
13+k(k+1), & 0\leq k<4,\\
\dfrac{k^{2}+7k+20}{2}, & k\geq4,
\end{cases}
\qquad\text{and}\qquad
\mu^{k}(f,g)=
\begin{cases}
16+k(k+1), & 0\leq k<4,\\
\dfrac{k^{2}+9k+20}{2}, & k\geq4.
\end{cases} μ k ( f g ) = ⎩ ⎨ ⎧ 13 + k ( k + 1 ) , 2 k 2 + 7 k + 20 , 0 ≤ k < 4 , k ≥ 4 , and μ k ( f , g ) = ⎩ ⎨ ⎧ 16 + k ( k + 1 ) , 2 k 2 + 9 k + 20 , 0 ≤ k < 4 , k ≥ 4.
Some values:
Table 1.
(3) Let f = x y + x 3 f=xy+x^{3} f = x y + x 3 and g = x y + x 3 + x 4 + y 5 g=xy+x^{3}+x^{4}+y^{5} g = x y + x 3 + x 4 + y 5 , so m = ℓ = 2 m=\ell=2 m = ℓ = 2 . Here μ ( f g ) = μ ( f ) + μ ( g ) + 2 I 0 ( f , g ) − 1 = 19 \mu(fg)=\mu(f)+\mu(g)+2I_{0}(f,g)-1=19 μ ( f g ) = μ ( f ) + μ ( g ) + 2 I 0 ( f , g ) − 1 = 19 , since μ ( f ) = μ ( g ) = 1 \mu(f)=\mu(g)=1 μ ( f ) = μ ( g ) = 1 and I 0 ( f , g ) = 9 I_{0}(f,g)=9 I 0 ( f , g ) = 9 . The member α f + β g = ( α + β ) ( x y + x 3 ) + β ( x 4 + y 5 ) \alpha f+\beta g=(\alpha+\beta)(xy+x^{3})+\beta(x^{4}+y^{5}) α f + β g = ( α + β ) ( x y + x 3 ) + β ( x 4 + y 5 ) is a node unless α + β = 0 \alpha+\beta=0 α + β = 0 . The fibers of the pencil are nodes, except h = f − g = − ( x 4 + y 5 ) h=f-g=-(x^{4}+y^{5}) h = f − g = − ( x 4 + y 5 ) , with μ ( h ) = 12 \mu(h)=12 μ ( h ) = 12 . Hence B ∗ ( f , g ) = { [ 1 : − 1 ] } \mathcal{B}^{*}(f,g)=\{[1:-1]\} B ∗ ( f , g ) = {[ 1 : − 1 ]} , μ g e n = 1 \mu_{\mathrm{gen}}=1 μ gen = 1 and Δ = 12 − 1 = 11 \Delta=12-1=11 Δ = 12 − 1 = 11 . Since ν ( h ) = 4 \nu(h)=4 ν ( h ) = 4 and ν g e n = 2 \nu_{\mathrm{gen}}=2 ν gen = 2 , we have δ = 2 \delta=2 δ = 2 , and Proposition 6.1 gives m 0 = m + ℓ + 1 = 5 m_{0}=m+\ell+1=5 m 0 = m + ℓ + 1 = 5 . Indeed, the initial forms of P P P and Q Q Q are − 3 x 4 y -3x^{4}y − 3 x 4 y and x 5 x^{5} x 5 , respectively. By Theorem 6.2 ,
μ k ( f , g ) = { μ k ( f g ) + 11 , 0 ≤ k < 4 , μ k ( f g ) + 12 , 4 ≤ k < 5 , μ k ( f g ) + 2 k + 4 , k ≥ 5. \mu^{k}(f,g)=
\begin{cases}
\mu^{k}(fg)+11, & 0\leq k<4,\\
\mu^{k}(fg)+12, & 4\leq k<5,\\
\mu^{k}(fg)+2k+4, & k\geq5.
\end{cases} μ k ( f , g ) = ⎩ ⎨ ⎧ μ k ( f g ) + 11 , μ k ( f g ) + 12 , μ k ( f g ) + 2 k + 4 , 0 ≤ k < 4 , 4 ≤ k < 5 , k ≥ 5.
where
μ k ( f g ) = { 19 + k ( k + 1 ) , 0 ≤ k < 4 , k 2 + 7 k + 32 2 , k ≥ 4 , and μ k ( f , g ) = { 30 + k ( k + 1 ) , 0 ≤ k < 5 , k 2 + 11 k + 40 2 , k ≥ 5. \mu^{k}(fg)=
\begin{cases}
19+k(k+1), & 0\le k<4,\\
\frac{k^{2}+7k+32}{2}, & k\ge4,
\end{cases}
\qquad\text{and}\qquad
\mu^{k}(f,g)=
\begin{cases}
30+k(k+1), & 0\le k<5,\\
\frac{k^{2}+11k+40}{2}, & k\ge5.
\end{cases} μ k ( f g ) = { 19 + k ( k + 1 ) , 2 k 2 + 7 k + 32 , 0 ≤ k < 4 , k ≥ 4 , and μ k ( f , g ) = { 30 + k ( k + 1 ) , 2 k 2 + 11 k + 40 , 0 ≤ k < 5 , k ≥ 5.
In [3 ] and [25 ] , a meromorphic germ ( f : g ) : ( C 2 , 0 ) ⇢ P 1 (f:g):(\mathbb{C}^{2},0)\dashrightarrow\mathbb{P}^{1} ( f : g ) : ( C 2 , 0 ) ⇢ P 1 , with f , g ∈ m f,g\in\mathfrak{m} f , g ∈ m relatively prime and μ ( f , g ) < ∞ \mu(f,g)<\infty μ ( f , g ) < ∞ , is called semitame if its only possible bifurcation fibers are f f f and g g g , that is, B ∗ ( f , g ) = ∅ \mathcal{B}^{*}(f,g)=\varnothing B ∗ ( f , g ) = ∅ . As a consequence of Proposition 6.1 and Theorem 6.2 , the following corollary characterizes semitame pairs in terms of the k k k -th Milnor number.
Corollary 6.5. Let f , g ∈ m ⊂ O 2 f,g\in\mathfrak{m}\subset\mathcal{O}_{2} f , g ∈ m ⊂ O 2 be relatively prime reduced germs such that μ ( f , g ) < ∞ \mu(f,g)<\infty μ ( f , g ) < ∞ . Then μ k ( f , g ) ≥ μ k ( f g ) \mu^{k}(f,g)\ge\mu^{k}(fg) μ k ( f , g ) ≥ μ k ( f g ) for every integer k ≥ 0 k\ge0 k ≥ 0 , and the following conditions are equivalent:
( f : g ) (f:g) ( f : g ) is semitame;
μ k ( f , g ) = μ k ( f g ) \mu^{k}(f,g)=\mu^{k}(fg) μ k ( f , g ) = μ k ( f g ) for every integer k ≥ 0 k\ge0 k ≥ 0 ;
μ k ( f , g ) = μ k ( f g ) \mu^{k}(f,g)=\mu^{k}(fg) μ k ( f , g ) = μ k ( f g ) for some integer k ≥ 0 k\ge0 k ≥ 0 .
Proof. For a fixed k k k , ε k ( a ) \varepsilon_{k}(a) ε k ( a ) is a non-increasing function of a a a . Since δ ≥ 0 \delta\ge0 δ ≥ 0 , Theorem 6.2 gives μ k ( f , g ) − μ k ( f g ) = Δ + ε k ( M − 1 ) − ε k ( M − 1 + δ ) ≥ Δ ≥ 0 \mu^{k}(f,g)-\mu^{k}(fg)=\Delta+\varepsilon_{k}(M-1)-\varepsilon_{k}(M-1+\delta)\ge\Delta\ge0 μ k ( f , g ) − μ k ( f g ) = Δ + ε k ( M − 1 ) − ε k ( M − 1 + δ ) ≥ Δ ≥ 0 for every k ≥ 0 k\ge0 k ≥ 0 , where every summand of Δ \Delta Δ is positive. This proves the inequality.
Assume first that ( f : g ) (f:g) ( f : g ) is semitame, that is, B ∗ ( f , g ) = ∅ \mathcal{B}^{*}(f,g)=\varnothing B ∗ ( f , g ) = ∅ . Then Δ = δ = 0 \Delta=\delta=0 Δ = δ = 0 , and the formula above gives μ k ( f , g ) = μ k ( f g ) \mu^{k}(f,g)=\mu^{k}(fg) μ k ( f , g ) = μ k ( f g ) for every k ≥ 0 k\ge0 k ≥ 0 , which is the second condition. The second condition clearly implies the third. Finally, suppose that μ k ( f , g ) = μ k ( f g ) \mu^{k}(f,g)=\mu^{k}(fg) μ k ( f , g ) = μ k ( f g ) for some k k k . By the inequality above, Δ ≤ 0 \Delta\le0 Δ ≤ 0 , so Δ = 0 \Delta=0 Δ = 0 . Since every summand of Δ \Delta Δ is positive, B ∗ ( f , g ) = ∅ \mathcal{B}^{*}(f,g)=\varnothing B ∗ ( f , g ) = ∅ , that is, ( f : g ) (f:g) ( f : g ) is semitame. Hence the three conditions are equivalent. □
Remark 6.6. By Proposition 6.1 , if ( f : g ) (f:g) ( f : g ) is semitame then ν ( F f / g ) = ν ( f ) + ν ( g ) − 1 \nu(\mathcal{F}_{f/g})=\nu(f)+\nu(g)-1 ν ( F f / g ) = ν ( f ) + ν ( g ) − 1 ; that is, the foliation F f / g \mathcal{F}_{f/g} F f / g has the smallest possible algebraic multiplicity. The converse does not hold, as Example 6.4 (1) shows.
Example 6.7. Let f = x 2 + y 3 f=x^{2}+y^{3} f = x 2 + y 3 and g = x 3 + y 2 g=x^{3}+y^{2} g = x 3 + y 2 . Here m = ℓ = 2 m=\ell=2 m = ℓ = 2 and m 0 = 3 = m + ℓ − 1 m_{0}=3=m+\ell-1 m 0 = 3 = m + ℓ − 1 , since the initial forms of P = g f x − f g x P=gf_{x}-fg_{x} P = g f x − f g x and Q = g f y − f g y Q=gf_{y}-fg_{y} Q = g f y − f g y are 2 x y 2 2xy^{2} 2 x y 2 and − 2 x 2 y -2x^{2}y − 2 x 2 y , respectively. The member α f + β g \alpha f+\beta g α f + β g with α β ≠ 0 \alpha\beta\ne0 α β = 0 has quadratic part α x 2 + β y 2 \alpha x^{2}+\beta y^{2} α x 2 + β y 2 , so it is a node. Hence μ g e n = 1 \mu_{\mathrm{gen}}=1 μ gen = 1 , B ∗ ( f , g ) = ∅ \mathcal{B}^{*}(f,g)=\varnothing B ∗ ( f , g ) = ∅ , that is, ( f : g ) (f:g) ( f : g ) is semitame, and Δ = 0 \Delta=0 Δ = 0 . Since f f f and g g g are cusps with I 0 ( f , g ) = 4 I_{0}(f,g)=4 I 0 ( f , g ) = 4 , we have μ ( f , g ) = μ ( f g ) = 2 + 2 + 2 ⋅ 4 − 1 = 11 \mu(f,g)=\mu(fg)=2+2+2\cdot4-1=11 μ ( f , g ) = μ ( f g ) = 2 + 2 + 2 ⋅ 4 − 1 = 11 . By Corollary 6.5 ,
μ k ( f , g ) = μ k ( f g ) = { 11 + k ( k + 1 ) , 0 ≤ k < 3 , 11 + k ( k + 1 ) − ( k − 2 ) ( k − 3 ) 2 , k ≥ 3. \mu^{k}(f,g)=\mu^{k}(fg)=
\begin{cases}
11+k(k+1), & 0\le k<3,\\
11+k(k+1)-\frac{(k-2)(k-3)}{2}, & k\ge3.
\end{cases} μ k ( f , g ) = μ k ( f g ) = { 11 + k ( k + 1 ) , 11 + k ( k + 1 ) − 2 ( k − 2 ) ( k − 3 ) , 0 ≤ k < 3 , k ≥ 3.
Acknowledgments: Arturo Fernández-Pérez and Nancy Saravia-Molina would like to thank the Universidad de La Laguna, where this work was completed, for its hospitality.
Funding: The first author acknowledges support from CNPq Projeto Universal 408687/2023-1 “Geometria das Equações Diferenciais Algébricas” and from CNPq-Brazil PQ-306011/2023-9. The second author is supported by the Spanish grant PID2023-149508NB-I00, funded by MICIU/AEI/10.13039/501100011033 and by FEDER, UE. The third author received financial support from CAPES, Brazil. The first and fourth authors were also supported by the Vicerrectorado de Investigación (VRI) of the Pontificia Universidad Católica del Perú (PUCP) through grant DFI-2025-PI1275.
Data Availability Statement: Data sharing is not applicable to this article, as no datasets were generated or analyzed during the current study.
Conflict of Interest: The authors declare that they have no conflict of interest.
References
(Arturo Fernández Pérez) Department of Mathematics. Federal University of Minas Gerais. Av. Pres. Antônio Carlos, 6627 CEP 31270-901, Pampulha - Belo Horizonte - Brazil. ORCID: 0000-0002-5827-8828 Email address: fernandez@ufmg.br
(Evelia R. García Barroso) Departamento de Matemáticas, Estadística e Investigación Operativa, Instituto Universitario de Matemáticas y Aplicaciones (IMAULL). Universidad de La Laguna. Apartado de Correos 456. 38200 La Laguna, Tenerife, Spain. ORCID: 0000-0001-7575-2619 Email address: ergarcia@ull.es
(Marcela Ribeiro) Department of General Education, Universidade Federal de Ciência e Inovação de Minas Gerais – Campus Curvelo. Rua Raymundo Mattoso, 900 Santa Rita - Curvelo - MG 35790-636, Brazil. ORCID: 0009-0009-2047-6341 Email address: marcelaribeiro@cefetmg.br
(Nancy Saravia-Molina) Sección de Matemáticas, Departamento Académico de Ciencias, Pontificia Universidad Católica del Perú, Av. Universitaria 1801, San Miguel, Lima 32, Peru. ORCID: 0000-0002-2819-8835 Email address: nsaraviam@pucp.edu.pe