twistorial structure of loop quantum gravity transition...
TRANSCRIPT
Twistorial structure of loop quantum gravity transition amplitudes
Simone Speziale
Centre de Physique Theorique de Luminy, Marseille, France
ILQG 13-11-2012mainly based on work with Wolfgang Wieland 1207.6348
Outline and goals
Classical Theory
‚ Description of the covariant phase space in terms of twistorsalgebra of the area matching and simplicity constraints
‚ Smearing the connection-tetrad algebra to the holonomy-flux algebra before or aftersolving the simplicity constraints is equivalent: T˚SUp2q with AB holonomy
‚ Notion of simple twistors solving the simplicity constraints
Quantum theory
‚ Hilbert space represented via homogeneous functions on spinor space(instead of cylindrical functions)
‚ Dynamics as integrals in twistor space
‚ Embedding of the Regge data of the EPRL asymptotics in the initial phase space
Work and ideas shared with a number of collaboratorsL. Freidel, E. Livine, W. Wieland, B. Dittrich, C. Rovelli, M. Dupuis, J. Tambornino, . . .
This talk: w Wolfgang Wieland 1207.6348 PRD12 w Miklos Langvik (to appear)
Speziale — Twistors and LQG 2/24
Smearing and simplicity constraints
Phase spaces:
SmearingSLp2,Cq variables ùñ T˚SLp2,Cq
ó Solve simplicity constraints ó
(primary and secondary)
Ashtekar-Barbero variables ùñ T˚SUp2q
Does the diagram commute?
LQG path: ë spin foam path: the other one
In the continuum:
‚ primary: simple bivectors, unique metric structure
‚ secondary: embedding of AB variables in covariant phase space: Γ “ ΓpEq
In the discrete:
‚ primary: simple twistors, unique (twisted) geometry
‚ secondary: non-trivial embedding of T˚SUp2q in T˚SLp2,Cq. Discrete ΓpEq ?
Speziale — Twistors and LQG 3/24
Smearing and simplicity constraints
Phase spaces:
SmearingSLp2,Cq variables ùñ T˚SLp2,Cq
ó Solve simplicity constraints ó
(primary and secondary)
Ashtekar-Barbero variables ùñ T˚SUp2q
Strategy: do symplectic reduction of the primary constraints, and think of thesecondary constraints as providing a non trivial gauge-fixing section.
ñ allows to show commutativity of the procedure, while postponingknowning explicitly the form and solution of the secondary constraints
Technique: twistorial tools: parametrize the phase spaces and the constraints interms of twistors
ñ allows to embed the non-linear HF phase space into a linear one,explicit calculations, plus direct link with variables used in the twistedgeometry parametrization
Speziale — Twistors and LQG 4/24
Panorama
twistors Ñ covariant ðñ covariantC-area matching twisted geometry loop gravity
SLp2,CqÓ simplicity constraints Ó Ó
SU(2) spinors Ñ “open” twisted geometry ðñ loop gravityarea matching SUp2q
Ó closure Ó Gauss law
“closed” twisted geometries ðñ gauge-inv.(polyhedra) loop gravity
Ó shape matching
Regge calculus
Speziale — Twistors and LQG 5/24
LQG and twisted geometries
On a fixed graph HΓ “ L2rSUp2qLs PΓ “ T˚SUp2qL
represents a truncation of the theory to a finite number of degrees of freedom w Rovelli, PRD 10
These can be interpreted as discrete geometries, called twisted geometries w Freidel, PRD 10, . . .
X, gùñ
Npζq, j, ξ, Npζq
hol-flux area-angles
Each classical holonomy-flux configuration on a fixed graph can be visualized as acollection of adjacent polyhedra with extrinsic curvature between them
‚ Each polyhedron is locally flat: curvature emerges at the hinges, as in Regge calculus
‚ They induce a discontinuous discrete metric: two neighbouring polyhedra areattached by faces with same area but different shape
‚ ξ: extrinsic geometry
‚ Realizes a link suggested by Immirzi, Smolin et al., with the subtlelty of discontinuity: Reggecalculus is too “rigid” to capture the kinematical degrees of freedom of a spin network
‚ The matching with Regge geometries becomes exact in 2+1 dimensions
The picture can be embedded in a SLp2,Cq formalism described by simple twistors
Speziale — Twistors and LQG 6/24
Twistors
‚ A pair of spinors, ω and π, chiral under Lorentz transformations,
Z “
ˆ
ωA
π 9A
˙
P T :“ C2ˆ C2˚
Twistor “complex helicity”: πAωA“ rπ|ωy SLp2,Cq-invariant
Twistor norm (helicity) s :“ Im`
rπ|ωy˘
SUp2, 2q invariant
‚ physical picture as a massless particle with a certain spin and momentum
‚ geometric picture via incidence relation ωA “ iXA 9Aπ 9A§ X a light ray in Minkowski space for a null twistor s “ 0§ X a congruence of light rays for a generic twistor
‚ can provide notion of non-linear graviton
We will see that twistors can be thought of as non-linear gravitons in a completelydifferent way than Penrose’s original one, as classical counterparts of LQG’s quantumgeometry
Speziale — Twistors and LQG 7/24
Representation of the Lorentz algebra
‚ C2 carries a representation of the Lorentz algebra (the celestial sphere)
‚ Similarly, there is a representation of the Lorentz algebra on twistor space T
ΠAB“
1
2pL` iKqAB “
1
2ωpAπBq
‚ We are interested in a representation of a larger Poissonian algebra, the one ofT˚SLp2,Cq: the holonomy-flux algebra of (covariant) loop quantum gravity
‚ This can be achieved on T2Q pZ, Zq,
ΠAB“
1
2ωpAπBq, ΠAB
“1
2ωpAπBq, hAB “
ωAπB ´ πAωB
?πω?πω
imposing a complex first-class constraint called complex area matching
C “ rπ|ωy ´ rπ|ωy “ 0,
C8{{C – T˚SLp2,Cq
That is, both Lorentz generators and holonomies can be expressed as simple functions ona space of two twistors, provided they have the same complex helicity
Speziale — Twistors and LQG 8/24
Twistor Poisson brackets
The symplectic structure simplifies in the larger space T2:
T2 C “ 0 T˚SLp2,CqtωA, ωBu “ 0 th, hu “ 0tπA, ω
Bu “ δBA ùñ tΠi, hu “ τ ih
tπA, πBu “ 0 tΠi,Πju “ εijkΠk
In particular, momenta commute.
16 twistors T2 Ñ T˚SLp2,Cq 12C-area matching C
Ó simplicity constraints Ó
8 simple twistors Ñ T˚SUp2q 6SU(2) spinors area matching C
Note: the complete system of constraints is reducible
Speziale — Twistors and LQG 9/24
Simplicity constraints
Gauge-fixed approach Pick a time direction nI (typically time gauge, nI “ p1, 0, 0, 0q)(cfr. with gauge-invariant approach e.g. Bianca and Jimmy)
‚ Allows to define SU(2) norm }ω}2 :“ δA 9AωAω
9A and xω|πy :“ δA 9Aω9AπA
‚ Primary simplicity constraints: matching of left and right geometries
K ` γL “ 0 ô Π “ eiθΠ, γ “ cotθ
2
and the same for tilded variables: Π “ eiθ ¯Π
‚ In terms of spinors:`
rπ|ωy “ R` iI˘
F1 “ R´ γI “ 0, F2 “ xω|πy “ 0, F2 “ 0
‚ On each link, pM,Fi, Fiq form a second class system: tF2, F2u ff 0which is further reducible: M and F1 imply F1.
3 indep. 1st class constraints`
C,D :“ F1 ` F1
˘
aut`
Cred P R, F1, F1
˘
4 2st class constraints F2, F2
Speziale — Twistors and LQG 10/24
Counting
3 indep. 1st class constraints`
C,D :“ F1 ` F1
˘
aut`
Cred P R, F1, F1
˘
4 2st class constraints F2, F2
16 twistors T2 Ñ T˚SLp2,Cq 12C-area matching C
Ó simplicity constraints Ó
8 simple twistors Ñ T˚SUp2q 6SU(2) spinors area matching C
Counting:
Right-Down: T2 C T˚SLp2,Cq D,F2, F2 T˚SUp2q
16 -2x2 12 -1x2-4 6
Down-Right: T2 F C4 Cred T˚SUp2q
16 -2x2-4 8 -1x2 6
Speziale — Twistors and LQG 11/24
Solution of the simplicity constraints
Focus first on half-link: pω, πq,`
F1 “ R´ γI, F2 “ xω|πy˘
‚ F1 “ 0 ñ rπ|ωy “ pγ ` iqj, j P R
‚ orbits: tF1, }ω}u “1
1`γ2 }ω}
‚ convenient to introduce a F1-gauge-invariant spinor
zA “?
2JωA
}ω}iγ`1, }z} :“
?2J, tF1, z
Au “ 0
constraint surface orbitspωA, πAq ÝÑ pωA, jq ÝÑ zA
Fi “ 0 }ω}
Symplectic reduction: C4{{F – C2
‚ Solutions parametrized by a SU(2) spinor zA
(transforms linearly under rotations but not under boosts)
‚ Simple twistors (on the 5d constraint surface): Z “ pωA, jq “ pzA, }ω}q
Speziale — Twistors and LQG 12/24
Area matching and T ˚SUp2q
On the link:16 pωA, πA, ωA, πAq P T2 Ñ pΠ, hq 12
C-area matching C
Ó simplicity constraints Ó
8 pzA, zAq P C4 Ñ pX, gq 6area matching Cred
Cred “ }z} ´ }z} C4{{Cred “ T˚SUp2q Q pX, gq Freidel PRD10
Symplectic reduction by the linear primary simplicity constraints selects an SUp2qholonomy-flux algebra from the covariant SLp2,Cq one
‚ Flux: X “ L “ ´K{γ‚ Holonomy X Ø h „ AγAB Ø A
Recall that Π “ ´hΠh´1. From this plus Fi “ Fi “ 0, it follows that
ph:hqABωB “ e´ΞωA, ph:hqABπ
B “ eΞπA
Using this one proves that the reduced g is the holonomy of the Ashtekar-Barberoconnection, a non-trivial and γ-dependent mixing of real and imaginary parts oflnh “ Γ` iK:
g „ eRe lnheγIm lnh ` opΞq
Speziale — Twistors and LQG 13/24
Geometric interpretation
Properties of the reduction:
‚ On C “ 0 surface, 1st class diagonal simplicity constraint D “ F1 ` F1
‚ Each orbit of D spanned by
Ξ :“ 2 ln
ˆ
}ω}
}ω}
˙
‚ In the time gauge,nIΛphq
IJnJ “ ´ cosh Ξ
The dihedral angle parametrizes the orbits of the diagonal simplicity constraint
‚ Symplectic reduction eliminates the dependence on Ξ, which is the coordinate of theorbits of the diagonal simplicity constraint D
‚ Secondary constraints: provide a non-trivial gauge-fixing of the orbits and thusnon-trivial embedding of T˚SUp2q in T˚SLp2,Cq, precisely as in the continuum
Speziale — Twistors and LQG 14/24
Geometric interpretation: covariant twisted geometries
‚ gauge-fixed approach pick a time direction nI (typically nI “ p1, 0, 0, 0q)
nIΛphqIJnJ “ ´ cosh Ξ, Ξ :“ 2 ln
ˆ
}ω}
}ω}
˙
pΠ, hq ÞÑ´
rπ|ωy, ζ, ζ, α, α, ξ,Ξ¯
given by
ζ “ ´ω1
ω0, α “
e2i argpω0q
rπ|ωyF2, ξ :“ 2 argpω0
q ´ 2 argpω0q ` γΞ.
reduction to twisted geometries:
Π “ ´ i2rπ|ωynpζqTατ3T
´1α n´1
pζq R´ γI “ α “ α “ 0 X “ jnpζqτ3n´1pζq
ÝÑ
h “ npζqTαep´ξ`pγ´iqΞqτ3T´1
α n´1pζq Ξ “ 0 g “ npζqe´ξτ3n´1
pζq
A “ Γ` iK “ AγAB ´ pγ ´ iqK
Speziale — Twistors and LQG 15/24
More on the brackets and the role of γ
Symplectic potential of T˚SLp2,Cq: pΠ, hq ÞÑ´
rπ|ωy “ R` iI, ζ, ζ, α, α, ξ,Ξ¯
Θ “ ´2Re´
TrrΠh´1dhs¯
“
“ pR´ γIqdΞ` Idξ ´iI
1` |ζ|2pζdζ ´ ζdζq ´
1
1` |ζ|2pαrπ|ωydζ ` αxω|πsdζq
´iI
1` |ζ|2pζd
¯ζ ´
¯ζdζq ´
1
1` |ζ|2prπ|ωyαd
¯ζ ` xω|πs ¯αdζq.
“Abelian” sector, described by the pR, I, ξ,Ξq, with Poisson brackets
tR,Ξu “ 1, tR, ξu “ γ, tI,Ξu “ 0, tI, ξu “ 1.
On the constraint surface, we recover the twisted geometry brackets, in particular
tj, ξu “ 1
where ξ :“ 2 argpω0q ´ 2 argpω0
q ` γΞ encodes the extrinsic curvature(But to be able to extract it I need to know explicitly the solution of the secondaryconstraints, i.e. the embedding in the covariant phase space)
Speziale — Twistors and LQG 16/24
Geometry of a simple twistor
‚ Incidence relation, ωA “ iXA 9Aπ 9A, X PM iff null twistor, s “ 0
‚ Simple twistors (Fi “ 0q:
ωA “1
rnA
9Aeiθ2 π 9A, r :“
ja
1` γ2
}ω}2
1. Not null, but isomorphic to null twistors:
Z ÞÑ Zγ “ pωA, eiθ{2π 9Aq, spZγq “ 0
Isomorphism depends on γ, reduces to the identity for γ “ 8
2. F2 “ xω|πy “ 0 aligns the spinors’ null poles to the time normal
`Ipωq `1
r2kIpπq “
?2}ω}2nI
A simple twistor is a γ-null twistor with a time-like direction picked up
The projective twistor space PN does not depend on the “scale” jwhereas it depends on |xω|πy|
Speziale — Twistors and LQG 17/24
Twistor networks
Zsl
Ztl
Zsm
Graph Γ decorated with a twistor on each half-link,
pT2qL{{CL – T˚SLp2,CqL
Quantization gives covariant spin networks ΨΓ,pρl,klq,In rGls,
with the covariant holonomy-flux algebra
‚ The standard Penrose interpretation is a collection of spinning massless particles
‚ A new interpretation in terms of geometries can be achieved thanks to thegauge-invariance at the nodes (Ψrh´1
splqGl htplqs “ ΨrGls):
ÿ
lPn
~JLl “ÿ
lPn
~JRl “ 0
Established in two steps:
1. each left- and right-handed sector defines a twisted geometry
2. simplicity constraints impose the matching between the two geometries
Speziale — Twistors and LQG 18/24
Further comments and questions
‚ Twistor space carries a representation of the larger group SU(2,2) of conformaltransformations of Minkowski: Non-trivial effect on a twisted geometry!See upcoming paper with Miklos Langvik
‚ Explicit form of the secondary constraints? Open problemAssume they correspond to a discretization of the continuum secondary constraints:
torsionlessness, Γ “ ΓpEq
Such a discrete Levi-Civita connection is well known in the context of Regge calculusCan it be constructed also without shape matching conditions? Yes!See brand new paper by Haggard-Rovelli-Vidotto-Wieland 1211.2166
‚ Open question: is there a consistent classical dynamics for twisted geometries, oronly for the subsector of Regge geometries?
‚ Does shape mismatch play a role in the dynamics?
‚ Can torsion be consistently encoded in ξ alone?
– end of classical part –
Speziale — Twistors and LQG 19/24
Quantization
Dirac quantization: first quantize auxiliary space, then impose constraints: simplicity onhalf-links, then area matching on full link
‚ Quantize initial twistorial phase space, a la Schrodinger:
“
πA, ωBs “ ´i~δBA , fpωq P L2
pC2, d4ωq´
ω “ ω, π “ ´i~ BBω
¯
“
ˆπA, ˆωB‰
“ i~δBA , fpπq
‚ Convenient basis homogeneous functions f pρ,kqpλωq “ λ´k´1`iρλk´1`iρf pρ,kqpωq(carry a unitary, infinite dimensional representation of the Lorentz group)
‚ Simplicity constraints
F1fpρ,kq
pωq “ 0 ñ ρ “ γk
M “ F2F2 Mf pρ,kqpωq “ 0 ñ k “ j
‚ Solution space fpγj,jqjm pωq “ }ω}2piγj´j´1q
xj,m|j, ωyPerelomov
(Not a function of the F1-reduced phase space, C2 Q zA. Half-density?d2ωf
pγj,jqjm is.)
‚ Area matching constraint´
Mfρ,k b f ρ,k¯
pω, πq “ 0 ñ pρ “ ρ, k “ kq
‚ Solution space Gpjqmmpω, πq :“ f
pγj,jqjm pωqf
pγj,jqjm pπq
Speziale — Twistors and LQG 20/24
From homogeneous functions to cylindrical functions
‚ Solution space Gpjqmmpω, πq :“ f
pγj,jqjm pωqf
pγj,jqjm pπq
(Morally Perelomov coherent states up to non-Lorentz-inv. norms }ω} and }π})
‚ carries a representation of the holonomy-flux algebra though harmonic oscillators§ Fluxes: Schwinger representation of SLp2,Cq§ Holonomy: ordering ambiguities, more natural operators twisted geometries ζ, ξ, etc.
‚ And the usual cylidrical functions? The relation is a kernel in twistor space:
@
Gpjqm,m
D
pgq :“ş
T2 dµpZ, Zq eiSBF pZ,Z,gqG
pjqm,mpω, πq “ µpjqD
pγj,jqjm jmpgq,
‚ It allows us to recover the EPRL model
Apgq “ÿ
j
µf pjqTrj
´
ź
wPf
Dpγj,jqpgwg´1w q
¯
, ZC “
ż
ź
v,τ
dgvτ
Speziale — Twistors and LQG 21/24
Dynamics as integrals in twistor space@
Gpjqm,m
D
pgq :“ş
T2 dµpZ, Zq eiSBF pZ,Z,gqG
pjqm,mpω, πq “ µpjqD
pβj,jqjm jmpgq,
Two ingredients:
1. Measureż
T2gf
dωA ^ dωA^ . . . δCpMq fpZ, Zq “
ż
T˚SLp2,Cqd3Π^ dHaarh^ c.c. fpΠ, hq
FP fixing of the complex area matching condition§ direct geometric proof of gauge-invariance§ Morally the same as the Gaussian version by Etera, Maite and Johannes
2. BF action in terms of spinors
SwrB,As “ ΠABrtwsh
BArBws ` c.c. “
`
gg´1˘A
BpωBπA ` π
BωAq ` c.c.
§ bilinear in the spinors§ mixes bulk and boundary elements
See paper for details of proof and explicit value of µpjq
Speziale — Twistors and LQG 22/24
What do we learn from this?@
Gpjqm,m
D
pgq :“ş
T2 dµpZ, Zq eiSBF pZ,Z,gqG
pjqm,mpω, πq “ µpjqD
pβj,jqjm jmpgq,
1. As a path integral in twistors, action gives trivial dynamics in phase space
@
Gpjqm,m
D
pgq “
ż
T2
dµpZ, Zq eiStotpZ,Z,gq
2. non-trivial dynamics emerges by an interplay between boundary phase spacevariables (the twistors) and bulk degrees of freedom (the bulk holonomies g)key “gluing” role played by saddle point conditions
3. 4-simplex asymptotics and Regge data: 3d dihedral angles fit in the phase space(the ζ’s), whereas areas j and extrinsic geometry ξ,Ξ are thrown away, butrecovered, in two different ways:
§ areas identified with quantum spins, and summed over in the definition of the model§ dihedral angles Ξ encoded in the bulk holonomies g which become the
Regge-Levi-Civita holonomies at the saddle point
4. Relaxing wedge flatness, we can define a curvature tensor and decompose it intoirreps: Weyl, Ricci and torsional parts
§ Still of Petrov type D, like in Regge calculus§ Additional torsional components carried by certain pieces of the holonomy identified
All of this points in the same direction: can we rewrite the model in a fully coherent way,where the phase space variables are treated on equal footing?Similar ideas have been expressed at length, Daniele, Etera, Maite, Valentin, Laurent, . . .
Speziale — Twistors and LQG 23/24
Conclusions and outlook
Can we express the dynamics as integrals over phase space? i.e.ř
j Ñş
djdξ(see Etera and Maite for an euclidean model of this)
in such a way that exponential of a suitable discretization of GR immediately emerges,not only at holonomy saddle point. This should be a discretization of GR:
1. in the first order formalism
2. without shape matching
Again, the key open question is the existence of a dynamics for twisted geometries
Remarks:
‚ A priori this is not necessary: correct semiclassical limit may also emerge from coarsegraining graphs, and not graph by graph But we need to find this out!
‚ Regardless, there should be many different discrete versions of GR, like there are ofgauge theories on a lattice. Why do we know only one version?
Longer term questions
‚ Can the formulation as integrals in twistor space improve explicit calculations?
‚ Any connections with twistors of the complete, infinitely-many-dofs LQG theory?
– finis terrae –
Speziale — Twistors and LQG 24/24