p-adic periods and local Langlands correspondences Laurent Fargues The archimedean case X = moduli space of rigidifed rank 2 variations of Hodge structures of weight (1, 0), (0, 1) The archimedean case X = moduli space of rigidifed rank 2 variations of Hodge structures of weight (1, 0), (0, 1) S = smooth complex analytic space X (S) = {(F, Fil(F ⊗R OS ), ρ)}/ ∼ where : I F local system of R-vector spaces of rank 2 on S The archimedean case X = moduli space of rigidifed rank 2 variations of Hodge structures of weight (1, 0), (0, 1) S = smooth complex analytic space X (S) = {(F, Fil(F ⊗R OS ), ρ)}/ ∼ where : I F local system of R-vector spaces of rank 2 on S I Fil(F ⊗R OS ) locally direct factor sub bundle of rank 1 The archimedean case X = moduli space of rigidifed rank 2 variations of Hodge structures of weight (1, 0), (0, 1) S = smooth complex analytic space X (S) = {(F, Fil(F ⊗R OS ), ρ)}/ ∼ where : I F local system of R-vector spaces of rank 2 on S I Fil(F ⊗R OS ) locally direct factor sub bundle of rank 1 ` ´ I ∀s ∈ S, Fs , Fil(F ⊗ OS )s defines a Hodge structure of weight |{z} | {z } R-v.s. of dim. 2 (1, 0), (0, 1) C-line in Fs ⊗R C The archimedean case X = moduli space of rigidifed rank 2 variations of Hodge structures of weight (1, 0), (0, 1) S = smooth complex analytic space X (S) = {(F, Fil(F ⊗R OS ), ρ)}/ ∼ where : I F local system of R-vector spaces of rank 2 on S I Fil(F ⊗R OS ) locally direct factor sub bundle of rank 1 ` ´ I ∀s ∈ S, Fs , Fil(F ⊗ OS )s defines a Hodge structure of weight |{z} | {z } R-v.s. of dim. 2 C-line in Fs ⊗R C (1, 0), (0, 1) → Fil ∩ Fil = (0) (Hodge condition) The archimedean case X = moduli space of rigidifed rank 2 variations of Hodge structures of weight (1, 0), (0, 1) S = smooth complex analytic space X (S) = {(F, Fil(F ⊗R OS ), ρ)}/ ∼ where : I F local system of R-vector spaces of rank 2 on S I Fil(F ⊗R OS ) locally direct factor sub bundle of rank 1 ` ´ I ∀s ∈ S, Fs , Fil(F ⊗ OS )s defines a Hodge structure of weight |{z} | {z } R-v.s. of dim. 2 I C-line in Fs ⊗R C (1, 0), (0, 1) → Fil ∩ Fil = (0) (Hodge condition) ∼ ρ : R2 −−→ F rigidification (kill the Betti monodromy) The archimedean case X = moduli space of rigidifed rank 2 variations of Hodge structures of weight (1, 0), (0, 1) S = smooth complex analytic space X (S) = {(F, Fil(F ⊗R OS ), ρ)}/ ∼ where : I F local system of R-vector spaces of rank 2 on S I Fil(F ⊗R OS ) locally direct factor sub bundle of rank 1 ` ´ I ∀s ∈ S, Fs , Fil(F ⊗ OS )s defines a Hodge structure of weight |{z} | {z } R-v.s. of dim. 2 I C-line in Fs ⊗R C (1, 0), (0, 1) → Fil ∩ Fil = (0) (Hodge condition) ∼ ρ : R2 −−→ F rigidification (kill the Betti monodromy) GL2 (R) via the action on ρ " X The archimedean case X = moduli space of rigidifed rank 2 variations of Hodge structures of weight (1, 0), (0, 1) S = smooth complex analytic space X (S) = {(F, Fil(F ⊗R OS ), ρ)}/ ∼ where : I F local system of R-vector spaces of rank 2 on S I Fil(F ⊗R OS ) locally direct factor sub bundle of rank 1 ` ´ I ∀s ∈ S, Fs , Fil(F ⊗ OS )s defines a Hodge structure of weight |{z} | {z } R-v.s. of dim. 2 I C-line in Fs ⊗R C (1, 0), (0, 1) → Fil ∩ Fil = (0) (Hodge condition) ∼ ρ : R2 −−→ F rigidification (kill the Betti monodromy) " GL2 (R) X via the action on ρ Period morphism : ∨ X ,→ (F, Fil, ρ) 7−→ X = P1C ` ´ ρ−1 Fil ⊂ F ⊗ OS −−→ OS2 Hodge structure 7−→ Hodge filtration ∼ The archimedean case The period morphism ∨ X ,→ X = P1C is an GL2 (R)-equivariant embedding. The archimedean case The period morphism ∨ X ,→ X = P1C is an GL2 (R)-equivariant embedding. Image described by Hodge condition : X = H± = C \ R ⊂ P1C The archimedean case The period morphism ∨ X ,→ X = P1C is an GL2 (R)-equivariant embedding. Image described by Hodge condition : X = H± = C \ R ⊂ P1C Uniformisation : Y (S) = {Elliptic curves /S}/ ∼ (groupı̈d of...) The archimedean case The period morphism ∨ X ,→ X = P1C is an GL2 (R)-equivariant embedding. Image described by Hodge condition : X = H± = C \ R ⊂ P1C Uniformisation : Y (S) = {Elliptic curves /S}/ ∼ (groupı̈d of...) ˘ ∼ e (S) = Elliptic curves π : E → S + ρ : Z2 −− Y → R 1 π∗ Z | {z } relative Betti coho e has an action of GL2 (Z) Via the action of GL2 (Z) on Z2 , Y ¯ /∼ The archimedean case The period morphism ∨ X ,→ X = P1C is an GL2 (R)-equivariant embedding. Image described by Hodge condition : X = H± = C \ R ⊂ P1C Uniformisation : Y (S) = {Elliptic curves /S}/ ∼ (groupı̈d of...) ˘ ∼ e (S) = Elliptic curves π : E → S + ρ : Z2 −− Y → R 1 π∗ Z | {z } relative Betti coho e has an action of GL2 (Z) Via the action of GL2 (Z) on Z2 , Y Hodge structure on R 1 π∗ R + rigidification ρ induces ∼ e −− Y →X Forget the rigidication : GL2 (Z)-invariant morphism e −→ Y Y Induces ˆ ˜ ˆ ˜ e = GL2 (Z)\H± Y ' GL2 (Z)\Y ¯ /∼ In general... Same picture for more general hermitian symmetric spaces/ Shimura varieties... Hodge filtration embedding ∨ X ,→ X ∨ where X = hermitian symmetric space, X = compact dual. Unformisation of Shimura variety : ` ´ ∼ G (Q)\ X × G (Af )/K −−→ ShK (C) Rapoport-Zink spaces (Serre-Tate, Lubin-Tate, Drinfeld, Rapoport-Zink) H = a p-divisible group over Fp For example H = A[p ∞ ] where A = abelian variety over Fp , or a direct factor of it Rapoport-Zink spaces (Serre-Tate, Lubin-Tate, Drinfeld, Rapoport-Zink) H = a p-divisible group over Fp For example H = A[p ∞ ] where A = abelian variety over Fp , or a direct factor of it c : W (Fp )-schemes where p is locally nilpotent M −→ Sets S 7−→ (H, ρ)/ ∼ I H = p-divisible group/S I rigidification ρ : H ×Fp (S mod p) −→ H ×S (S mod p) is a quasi-isogeny (virtual isogeny) Rapoport-Zink spaces Theorem (Rapoport-Zink) I c is representable by a formal scheme locally formally of finite type over M Spf(W (Fp )). → locally of the form Spf(W (Fp )Jx1 , . . . , xn Khy1 , . . . , ym i/Ideal ´ Rapoport-Zink spaces Theorem (Rapoport-Zink) I I c is representable by a formal scheme locally formally of finite type over M Spf(W (Fp )). cred (locally of finite type over Fp ) has The reduced special fiber M projective irreducible components. → locally of the form Spf(W (Fp )Jx1 , . . . , xn Khy1 , . . . , ym i/Ideal ´ Rapoport-Zink spaces Theorem (Rapoport-Zink) I I c is representable by a formal scheme locally formally of finite type over M Spf(W (Fp )). cred (locally of finite type over Fp ) has The reduced special fiber M projective irreducible components. → locally of the form Spf(W (Fp )Jx1 , . . . , xn Khy1 , . . . , ym i/Ideal ´ r Y ` ´× J = End H Q ' GLmi (Dλi ) p i=1 where (λi )i are the Dieudonné-Manin slopes wt. multiplicities (mi )i Dλ := division algebra with invariant λ̄ ∈ Q/Z c via its action on the rigidification ρ. Then J acts on M Rapoport-Zink tower can = generic fiber as a Berkovich analytic space/W (Fp )Q M W J Rapoport-Zink tower can = generic fiber as a Berkovich analytic space/W (Fp )Q M W J n := ht(H). ` can π1 M ´ " Tp (univ .deformation) = rank n étale Zp -local system can M Rapoport-Zink tower can = generic fiber as a Berkovich analytic space/W (Fp )Q M W J n := ht(H). ` can π1 M ´ " Tp (univ .deformation) = rank n étale Zp -local system can M → GLn (Qp )-equivariant tower of étale coverings J (MK )K ⊂GLn (Zp ) GLn (Zp ) | GLn (Qp ) can = MGL (Z ) M n p GLn (Zp )-pro Galois covering + action of GLn (Qp ) (Hecke correspondences) Period mapping (Katz, Gross-Hopkins, Rapoport-Zink) D(H)Qp = rational Dieudonné module of H 1 (Hcris ) Period mapping (Katz, Gross-Hopkins, Rapoport-Zink) D(H)Qp = rational Dieudonné module of H 1 (Hcris ) D = convergent isocrystal associated to the Dieudonné crystal of the universal deformation = locally free OM can -module of finite rank + Gauss-Manin connection ∇ on D Fil D = Hodge filtration Period mapping (Katz, Gross-Hopkins, Rapoport-Zink) 1 (Hcris ) D(H)Qp = rational Dieudonné module of H D = convergent isocrystal associated to the Dieudonné crystal of the universal deformation = locally free OM can -module of finite rank + Gauss-Manin connection ∇ on D Fil D = Hodge filtration Rigidification ρ induces rigidification ∼ ρ∗ : D(H)Qp ⊗W (Fp )Q OM −→ D can − p D(H)Qp = D∇=0 Then Hodge filtration induces a period morphism J can −−−→ F π:M where F = an analytic Grassmanian associated to D(H)Qp Period mapping I can → F is quasi-étale (Grothendieck-Messing deformation theory, π:M analog of Kodaira-Spencer def. theory) Period mapping I I can → F is quasi-étale (Grothendieck-Messing deformation theory, π:M analog of Kodaira-Spencer def. theory) can ) = ∂(F) = ∅ ⇒ π is étale But ∂(M Period mapping I can → F is quasi-étale (Grothendieck-Messing deformation theory, π:M analog of Kodaira-Spencer def. theory) can ) = ∂(F) = ∅ ⇒ π is étale But ∂(M I The morphism π is Hecke invariant I g / MGLn (Z)∩gGLn (Zp )g −1 ∼ NNN p p NNN p p N& xppp can can WWW g M g g M WWWW π π ggggg WWWW ggg WWWW WWW+ sggggggg F MGLn (Z)∩gGLn (Zp )g −1 and the fibers of π are the Hecke orbits ' GLn (Qp )/GLn (Zp ) Period mapping I can → F is quasi-étale (Grothendieck-Messing deformation theory, π:M analog of Kodaira-Spencer def. theory) can ) = ∂(F) = ∅ ⇒ π is étale But ∂(M I The morphism π is Hecke invariant I g / MGLn (Z)∩gGLn (Zp )g −1 ∼ NNN p p NNN p p N& xppp can can WWW g M g g M WWWW π π ggggg WWWW ggg WWWW WWW+ sggggggg F MGLn (Z)∩gGLn (Zp )g −1 and the fibers of π are the Hecke orbits ' GLn (Qp )/GLn (Zp ) I Set F a = Im(π) = open subset of F (admissible locus) Theorem (de Jong) can −→ F a is an étale covering in the sens of de Jong and the Zp -local π:M system given by Tp (univ .def .) descends to a Qp -local system in the sens of de Jong corresponding to Vp (univ .def .) on F a Theorem (de Jong) can −→ F a is an étale covering in the sens of de Jong and the Zp -local π:M system given by Tp (univ .def .) descends to a Qp -local system in the sens of de Jong corresponding to Vp (univ .def .) on F a ∼ can , after fixing a basis Znp −− → if x =geo. point of M → Tp (Hxuniv ) can , x) π1dJ (M π1dJ (F a , π(x)) / π1 (M can , x) / GLn (Z _ p) / GLn (Qp ) En résumé : « M∞ := lim MK » ←− K O M∞ GLn (Zp ) GLn (Qp ) can M rigidification of p-adic étale coho. π Fa / rigidification of p-adic cristalline coho. \ J | GLn (Qp ) Fontaine’s condition What is the image of the period mapping F a ⊂ F ? In the archimedean setting : given by Hodge condition → easily compute the ∨ open subset X ⊂ X Fontaine’s condition What is the image of the period mapping F a ⊂ F ? In the archimedean setting : given by Hodge condition → easily compute the ∨ open subset X ⊂ X There is a natural open subset, the weakly amissible open subset, F wa , defined by Fontaine’s weakly admissibility condition associated to the filtered isocrystal F a ⊂ F wa ⊂ F Fontaine’s condition What is the image of the period mapping F a ⊂ F ? In the archimedean setting : given by Hodge condition → easily compute the ∨ open subset X ⊂ X There is a natural open subset, the weakly amissible open subset, F wa , defined by Fontaine’s weakly admissibility condition associated to the filtered isocrystal F a ⊂ F wa ⊂ F I By Fontaine (and his disciples...) the classical points of F a and F wa coincide. But in general F a ( F wa (counterexample by Rapoport and Zink, other one by Hartl). Fontaine’s condition What is the image of the period mapping F a ⊂ F ? In the archimedean setting : given by Hodge condition → easily compute the ∨ open subset X ⊂ X There is a natural open subset, the weakly amissible open subset, F wa , defined by Fontaine’s weakly admissibility condition associated to the filtered isocrystal F a ⊂ F wa ⊂ F I I By Fontaine (and his disciples...) the classical points of F a and F wa coincide. But in general F a ( F wa (counterexample by Rapoport and Zink, other one by Hartl). In general Faltings has given a p-adic Hodge theoretic description of F a proving a necessary condition given by Fontaine is sufficient to be admissible, but it is quite abstract and difficult to compute... Fontaine’s condition What is the image of the period mapping F a ⊂ F ? In the archimedean setting : given by Hodge condition → easily compute the ∨ open subset X ⊂ X There is a natural open subset, the weakly amissible open subset, F wa , defined by Fontaine’s weakly admissibility condition associated to the filtered isocrystal F a ⊂ F wa ⊂ F I I I By Fontaine (and his disciples...) the classical points of F a and F wa coincide. But in general F a ( F wa (counterexample by Rapoport and Zink, other one by Hartl). In general Faltings has given a p-adic Hodge theoretic description of F a proving a necessary condition given by Fontaine is sufficient to be admissible, but it is quite abstract and difficult to compute... F wa has an interpretation in terms of semi-stability in the Harder-Narasimhan sens of a filtered isocrystal and even semi-stability in the sens of Mumford (Totaro). Fontaine’s condition What is the image of the period mapping F a ⊂ F ? In the archimedean setting : given by Hodge condition → easily compute the ∨ open subset X ⊂ X There is a natural open subset, the weakly amissible open subset, F wa , defined by Fontaine’s weakly admissibility condition associated to the filtered isocrystal F a ⊂ F wa ⊂ F I I I I By Fontaine (and his disciples...) the classical points of F a and F wa coincide. But in general F a ( F wa (counterexample by Rapoport and Zink, other one by Hartl). In general Faltings has given a p-adic Hodge theoretic description of F a proving a necessary condition given by Fontaine is sufficient to be admissible, but it is quite abstract and difficult to compute... F wa has an interpretation in terms of semi-stability in the Harder-Narasimhan sens of a filtered isocrystal and even semi-stability in the sens of Mumford (Totaro). F a has an interpretation in terms of semi-stability of a vector bundle on a « curve lying on F »(joint work with J.M. Fontaine) p-adic uniformization (Serre-Tate, Cherednik-Drinfeld, Varshavsky, Rapoport-Zink) Theorem (Rapoport-Zink) The spaces MK uniformize p-adically pieces of the p-adic analytic spaces associated to some Shimura varieties. p-adic uniformization (Serre-Tate, Cherednik-Drinfeld, Varshavsky, Rapoport-Zink) Theorem (Rapoport-Zink) The spaces MK uniformize p-adically pieces of the p-adic analytic spaces associated to some Shimura varieties. I In general there is no p-adic uniformization of the whole Shimura variety. Only case : Drinfeld case. p-adic uniformization (Serre-Tate, Cherednik-Drinfeld, Varshavsky, Rapoport-Zink) Theorem (Rapoport-Zink) The spaces MK uniformize p-adically pieces of the p-adic analytic spaces associated to some Shimura varieties. I In general there is no p-adic uniformization of the whole Shimura variety. Only case : Drinfeld case. I In the non-basic case (when J is not an inner form of G : non-isoclinic case) R.Z. theorem is non-satisfactory. It has been strenghtened by Harris-Taylor and in general by Mantovan (Igusa varieties theory). Local Langlands correspondences One can define R.Z. tower of spaces for more general reductive groups than GLn/Qp (not all). G = reductive group over Qp J = inner form of a Levi subgroup of the quasi-split inner form of G (MK )K ⊂G (Zp ) W J | G (Qp ) Local Langlands correspondences One can define R.Z. tower of spaces for more general reductive groups than GLn/Qp (not all). G = reductive group over Qp J = inner form of a Levi subgroup of the quasi-split inner form of G (MK )K ⊂G (Zp ) W | G (Qp ) J Consider lim −→ K ⊂G (Zp ) ˆ p , Q` ) | Hc• (MK ⊗C J×G (Qp )×WE WE = Weil group of the reflex field, [E : Qp ] < +∞ Action of J × G (Qp ) is smooth (Berkovich for the action of J) Local Langlands correspondences One can define R.Z. tower of spaces for more general reductive groups than GLn/Qp (not all). G = reductive group over Qp J = inner form of a Levi subgroup of the quasi-split inner form of G (MK )K ⊂G (Zp ) W | G (Qp ) J Consider lim −→ K ⊂G (Zp ) ˆ p , Q` ) | Hc• (MK ⊗C J×G (Qp )×WE WE = Weil group of the reflex field, [E : Qp ] < +∞ Action of J × G (Qp ) is smooth (Berkovich for the action of J) General conjecture (vague form) : this representations realize local Langlands correspondences between rep. of G (Qp ), J and WE Results : the Lubin-Tate case Lubin-Tate case : deform a 1-dimensional formal p-divisible group/Fp . × G = GLn/F , J = D1/n where [F : Qp ] < +∞, D1/n = div. alg. wt. invariant 1/n over F a n−1 π can = M B̊ −−→ Pn−1 Z In this case π is surjective (Lafaille, Gross-Hopkins) : F a = F wa = F = Pn−1 . Results : the Lubin-Tate case Lubin-Tate case : deform a 1-dimensional formal p-divisible group/Fp . × G = GLn/F , J = D1/n where [F : Qp ] < +∞, D1/n = div. alg. wt. invariant 1/n over F a n−1 π can = M B̊ −−→ Pn−1 Z In this case π is surjective (Lafaille, Gross-Hopkins) : F a = F wa = F = Pn−1 . × Rep(D1/n ) → Rep(GLn (F )) : Jacquet Langlands correspondence Rep(GLn (F )) ↔ Rep(WF ) : local Langlands correspondence (in fact we want to use those cohomology spaces to prove it exists...) Results : the Lubin-Tate case I Deligne and Carayol proved the result for GL2 . Carayol gave a general conjecture for the supercuspidal part in the GLn case (depending on a good étale cohomology theory for rigid analytic spaces...) Results : the Lubin-Tate case I Deligne and Carayol proved the result for GL2 . Carayol gave a general conjecture for the supercuspidal part in the GLn case (depending on a good étale cohomology theory for rigid analytic spaces...) I Berkovich developped his theory of étale cohomology of analytic spaces/formal vanishing cycles... Results : the Lubin-Tate case I Deligne and Carayol proved the result for GL2 . Carayol gave a general conjecture for the supercuspidal part in the GLn case (depending on a good étale cohomology theory for rigid analytic spaces...) I Berkovich developped his theory of étale cohomology of analytic spaces/formal vanishing cycles... I Boyer proved Carayol’s conjecture for the supercuspidal part in the equal characteristic case (Fq ((π))) (the local Langlands corr. in the equal char. case was allready knwon by Laumon-Rapoport-Stuhler) Results : the Lubin-Tate case I Deligne and Carayol proved the result for GL2 . Carayol gave a general conjecture for the supercuspidal part in the GLn case (depending on a good étale cohomology theory for rigid analytic spaces...) I Berkovich developped his theory of étale cohomology of analytic spaces/formal vanishing cycles... I Boyer proved Carayol’s conjecture for the supercuspidal part in the equal characteristic case (Fq ((π))) (the local Langlands corr. in the equal char. case was allready knwon by Laumon-Rapoport-Stuhler) I Harris and Taylor proved the local Langlands conjecture in the p-adic case, using those cohomology spaces and computed Pcompletly the alternate sum of the cohomology of the Lubin-Tate tower, i (−1)i [Hci ], as virtual representations in terms of the local Langlands correspondences Results : the Lubin-Tate case I Deligne and Carayol proved the result for GL2 . Carayol gave a general conjecture for the supercuspidal part in the GLn case (depending on a good étale cohomology theory for rigid analytic spaces...) I Berkovich developped his theory of étale cohomology of analytic spaces/formal vanishing cycles... I Boyer proved Carayol’s conjecture for the supercuspidal part in the equal characteristic case (Fq ((π))) (the local Langlands corr. in the equal char. case was allready knwon by Laumon-Rapoport-Stuhler) I Harris and Taylor proved the local Langlands conjecture in the p-adic case, using those cohomology spaces and computed Pcompletly the alternate sum of the cohomology of the Lubin-Tate tower, i (−1)i [Hci ], as virtual representations in terms of the local Langlands correspondences I Boyer was able to separate the terms in the alternate sum and computed each individual cohomology spaces in any degree without taking an alternate sum (uses in particular an amelioration of Berkovich’s invariance under formal completion of vanishing cycles (F.)) Results : the Lubin-Tate case I Deligne and Carayol proved the result for GL2 . Carayol gave a general conjecture for the supercuspidal part in the GLn case (depending on a good étale cohomology theory for rigid analytic spaces...) I Berkovich developped his theory of étale cohomology of analytic spaces/formal vanishing cycles... I Boyer proved Carayol’s conjecture for the supercuspidal part in the equal characteristic case (Fq ((π))) (the local Langlands corr. in the equal char. case was allready knwon by Laumon-Rapoport-Stuhler) I Harris and Taylor proved the local Langlands conjecture in the p-adic case, using those cohomology spaces and computed Pcompletly the alternate sum of the cohomology of the Lubin-Tate tower, i (−1)i [Hci ], as virtual representations in terms of the local Langlands correspondences I Boyer was able to separate the terms in the alternate sum and computed each individual cohomology spaces in any degree without taking an alternate sum (uses in particular an amelioration of Berkovich’s invariance under formal completion of vanishing cycles (F.)) I Dat computed the smooth-equivariant cohomology complex of the Lubin-Tate tower, using the result by Boyer Results : the Drinfled case × In this case switch the G and J of the Lubin-Tate case : G = D1/n and G = GLn/F a π can ' Ω −−→ Pn−1 M Z where Ω = Pn−1 \ [ H H∈P̌n−1 (F ) can , the period and F wa = F a = Ω : on each connected component of M morphism is an iso. onto its image. Results : the Drinfled case × In this case switch the G and J of the Lubin-Tate case : G = D1/n and G = GLn/F a π can ' Ω −−→ Pn−1 M Z where Ω = Pn−1 \ [ H H∈P̌n−1 (F ) can , the period and F wa = F a = Ω : on each connected component of M morphism is an iso. onto its image. Theorem (Faltings, F.) The cohomology of the Drinfeld tower is isomorphic to the one of the Lubin-Tate tower after switching G and J... Remark : in fact this theorem is used by Boyer and Dat in the preceding theorems about the L.T. tower. Other cases Theorem (F.) G = GLn/F with F |Qp unramified and J = Qp -points of an inner form of G equal either to the trivial inner form or an anisotropic inner form. Then the supercuspidal part of the cohomology of the R.Z. tower realizes local Langlands correspondences. Same type of result for a p-adic unitary group in three variables Other cases Theorem (F.) G = GLn/F with F |Qp unramified and J = Qp -points of an inner form of G equal either to the trivial inner form or an anisotropic inner form. Then the supercuspidal part of the cohomology of the R.Z. tower realizes local Langlands correspondences. Same type of result for a p-adic unitary group in three variables Restrictions on the groups : lack of results in harmonic analysis at that time, not on the geometric part of the proof. Since this theorem (2001), the technics of harmonic analysis have evolved and it is quite probable one can use my technic of proof to prove the supercuspidal part of any basic (J is an inner form of G ) R.Z. space (linear group, symplectic groups, unitary groups) realizes local Langlands correspondences.