Fortschritte der Physik, 27 May 2022

MPP-2010-21
Basics of F-theory from the Type IIB Perspective
Ralph Blumenhagen1,
1 Max-Planck-Institut für Physik, Föhringer Ring 6, 80805 München, Germany

These short lecture notes provide an introduction to some basic notions of F-theory with some special
emphasis on its relation to Type IIB orientifolds with O7/O3-planes.
arXiv:1002.2836v1 [hep-th] 15 Feb 2010

Copyright line will be provided by the publisher

1 Introduction
Historically outstripping heterotic string compactifications and intersecting D-brane models, the last two
years have seen the main activity in the field of string phenomenology shifting towards F-theory models. In
this framework some of the model building shortcomings of D-brane realizations of grand unified theories
(GUTs) can be overcome.
To appreciate this, let us recall some issues on D-brane constructions. The gauge theories are sup-
ported on D-branes, which in general can have a dimension smaller than the bulk. Completely occupying
our observable large scale four-dimensional world, they wrap certain sub-manifolds of the internal ge-
ometry. The matter fields are localized on the intersections of such D-branes. Since the early years of
so-called intersecting D-brane models, it was clear that this set-up naturally allows for semi-simple gauge
groups with matter fields in the bifundamental representations. Therefore, here one directly engineers
the SU(3)c × SU(2)w × U(1)Y MSSM, while leaving the unification of gauge couplings at the GUT scale
essentially unexplained. Indeed, the gauge couplings depend on the generally different volumes of the
internal cycles wrapped by the D-branes supporting each gauge factor.
It was quickly realized that the construction of GUT groups SO(10) and SU(5) was obstructed by the
perturbative absence of matter fields in the 16 representation of SO(10) respectively by the absence of
the top Yukawa coupling 10 10 5H for the SU(5) case. These two latter features are of non-perturbative
origin for orientifold models (see [1] for a review). It was realized two years ago that the aforementioned
problems with realizing simple GUT groups in orientifold constructions are nicely reconciled in F-theory
models on elliptically fibered Calabi-Yau four-folds [2–5] (see [6] for a more phenomenological review).
One can think of F-theory as Type IIB compactifications on compact complex n-dimensional manifolds
Bn with general (p, q) 7-branes wrapping 2(n − 1) cycles of Bn . Since the 7-branes are of real co-dimension
two, the solutions to the Laplace equations are of logarithmic type. Therefore, the backreaction of the
7-branes on the geometry and the dilaton is always substantial and has to be taken into account. By
identifying the strong-weak SL(2, ZZ) duality of the Type IIB superstring with the modular group of a
torus, C. Vafa [7] showed that the backreaction can geometrically be taken into account by an elliptic
fibration over the base Bn , where the modular parameter of the fiber is identified with the axio-dilaton field
of Type IIB. The location of the 7-branes correspond to the degeneration loci of the elliptic fibration and
for supersymmetry the fibrations have to be of Calabi-Yau type.
Due to the strong backreaction, only in a global gs → 0 limit a general F-theory model is expected
to correspond to an orientifold. F-theory inherently contains some features which are non-perturbative
from the orientifold point of view. This is the reason for the appearance of exceptional groups in F-theory,
which by a further breaking also realize the spinor representation of a GUT SO(10) as well as the top-quark
Yukawa couplings 10 10 5H in GUT SU(5). For four-dimensional models, the basis B3 is a Fano three-fold
and the 7-branes wrap complex surfaces, i.e. four-cycle.
Thus, F-theory is a non-perturbative completion of Type IIB orientifolds where the 7-branes are com-
pletely encoded in the geometry of the elliptic fibration. The aim of this lecture is to give an introduction
Copyright line will be provided by the publisher
2 Ralph Blumenhagen: Basics of F-theory

into some basis notions of F-theory, which essentially addresses those students, who are already familiar
with D-brane constructions. It is explained why F-theory is inevitable for the correct study of Type IIB
compactifications with D7-branes, in which sense it goes beyond the perturbative Type IIB superstring and
how this leads to a solution of the above mentioned problems with GUT models. Note that this was the first
of a series of two lectures on F-theory GUTs held at the 9th Hellenic School on Elementary Particle Physics
and Gravity, Corfu 2009. The second lecture focused more on the specifics of realizing four-dimensional
SU(5) GUTs from F-theory. Please consult [8] for more detailed lecture notes on F-theory.

2 D7-branes and SL(2, ZZ) self-duality

As a starting point, we consider Type IIB orientifolds compactified on a Calabi-Yau three-fold X and an
orientifold projection Ωσ (−1)FL (see [9] for a review). Here σ denotes a holomorphic involution of X
acting as

σ ∗ (J) = J, σ ∗ (Ω3 ) = −Ω3 (1)

on the Kähler respectively holomorphic (3, 0)-form of X. This orientifold quotient introduces an O7-plane
into the theory, whose tadpole is canceled by the introduction of stacks of D7-branes wrapping various
holomorphic four-cycles of X, whose total homology class in H4 (X, ZZ) is equal to the one of the O7-
plane. One can now compute the (chiral) massless spectrum coming from the lowest excitations of open
strings stretched between various pairs of D7-branes. This gives rise to gauge bosons of only unitary or
orthogonal/symplectic gauge groups and in addition to matter fields transforming solely in bifundamental
or (anti)-symmetric representations of the gauge group. This is simply a consequence of the fact that an
open string has two ends. Clearly, such open string excitations can never give rise to exceptional gauge
groups and, as a group theoretic consequence, to matter in the spinor representation of an SO(10) gauge
group.
What we have just briefly described is the construction of perturbative Type IIB orientifold string vacua
and its short-comings when it comes to GUT like structures. Naively, this seems to be the end of the story.
However, taking the perturbative string limit gs ≪ 1 is, to say the least, quite questionable, if branes of
(real) co-dimension two, such as D7-branes, are present. This becomes evident by studying the D7-brane
solution in Type IIB supergravity [10], which is magnetically charged under the R-R scalar field C0 with
corresponding field strength F1 = dC0 .
The space transverse to the D7-brane is two-dimensional so that it is convenient to combine the two
transverse coordinates into a single complex variable z = y1 + iy2 . Furthermore ones combine the Type
IIB dilaton together with the R-R scalar field C0 into a complex scalar field τ = C0 + ie−Φ . For the D7-
brane supergravity solution, τ will be a function of the complex coordinate z, and the field equation, i.e.
the Laplace equation in two-dimensions, is now written as ∂z̄ τ (z, z̄) = 0. This means that τ must be a
holomorphic function. However not any holomorphic solution is a good solution, e.g. we must require
Imτ > 0. Furthermore the solution must have finite energy per unit volume, and it turns out that for this
purpose one needs the SL(2, ZZ) action on τ . The solution can then be written as j(τ ) = z−1 , where

3
ϑ38 (τ ) + ϑ48 (τ ) + ϑ28 (τ )
j(τ ) = = e−2π iτ + 744 + 196884 e2π iτ + . . . , (2)
8 η 24 (τ )

is the modular invariant j-function. Close to the D7-brane, i.e. at |z| → 0 this solution behaves as
1
τ (z) ∼ log z , (3)
2π i
which, circling once around the origin, gives rise to a monodromy τ → τ + 1. This monodromy reflects
that the D7-brane carries C0 charge one. Due to this logarithmic dependence of the axio-dilaton on the
transverse coordinate z, the backreaction of the D7-brane is so strong that one cannot really control the
Copyright line will be provided by the publisher
fdp header will be provided by the publisher 3

weak-coupling regime. Once the string coupling is non-zero and maybe small somewhere, it necessarily
becomes large in other regions of the transverse space. However, eq.(3) implies that gs = exp(Φ) is small
close to the D7-brane so that for the gauge theory on the D7-brane we expect a weak coupling description.
As will now review, also the backreaction on the metric is strong. Consider the 10-dimensional space-
time metric of the D7-brane solution: ds2 = −dt 2 + ∑7i=1 dx2i + eB(z,z̄) dz dz̄. Then the Einstein equation
connects the warp factor B with the dilaton field, and one obtains the following simple relation ∂ ∂¯ B =
∂ ∂¯ log(Imτ ). It turns out that the solution with the correct modular properties and the right asymptotic
behavior far away from the D7-brane is given by
η 2 (τ ) η̄ 2 (τ̄ )
eB(z,z) = (Imτ ) 1 . (4)
∏Ni=1 (z − zi ) 6

Here N is the number of 7-branes and the zi denote their positions in the two-dimensional transverse space.
Expanding this function for large |z| one gets that B(z, z) ∼ −N/12 · log|z|. Using this asymptotic behavior
one realizes that the metric goes like |z−N/12 dz|2 far away from the 7-branes. This means that each 7-brane
leaves a deficit angle of 2π /12 in the transverse space. In fact precisely 24 7-branes are required to get
the deficit angle 4π of the compact two-dimensional sphere CIP1 . Therefore, there exists a supersymmetric
compactification of the Type IIB superstring on CIP1 with precisely 24 7-branes and a varying dilaton.
So far we have only been talking about D7-branes. However, due to the SL(2, ZZ) duality symmetry
there exist infinitely many different kinds of 7-branes. Indeed, since there exists a doublet of two-forms
(B2 ,C2 ) in the ten-dimensional Type IIB string theory, there are not only fundamental strings and D1-
branes, but also strings carrying electric charges (p, q), where in this notation a fundamental string is a
(1, 0) string. Acting with an SL(2, ZZ) transformation on the fundamental string gives
    
p p r 1
= with ps − qr = 1 . (5)
q q s 0

Similarly, one has a doublet of ten-dimensional scalars (C0 , Φ) leading to magnetically charged (p, q) 7-
branes, where a D7-brane (charged only under C0 ) is a (1, 0) 7-brane. Since a fundamental string can
end on a D7-brane, SL(2, ZZ) implies that a (p, q) string can end on a (p, q) 7-brane. In eq.

(3) we have
seen that the solitonic solution of a D7-brane induces an SL(2, ZZ) monodromy MD7 = 10 11 . Applying the
SL(2, ZZ) symmetry, a general (p, q) 7-brane induces a monodromy
−1 
p2
   
p r 1 1 p r 1 − pq
M(p,q) = = . (6)
q s 0 1 q s −q2 1 + pq

For later purpose we consider the three 7-branes A = (1, 0), B = (1, −1) and C = (1, 1). It is now straight-
forward to compute the monodromy matrices for the combinations of these three 7-branes listed in table
1. To understand the massless modes between such more general (p, q) 7-branes, one notices that the
(p, q)-strings can form so-called string junctions. For instance a (1, 1) string can split into a (1, 0) and
(0, 1) string. Similar to open (1, 0) strings ending on D7-branes, for more general 7-brane there can exist
so-called string junctions ending on them. For instance there can be a string junction with four external
strings of type (1, 0) − (1, 0) − (1, 1) − (1, −1), which can end on the respective 7-branes A − A − B − C.
Clearly, such objects can give qualitatively new massless states beyond what is possible with perturbative
fundamental open strings. We will come back to this in the next section.

3 F-theory
The observations made in the previous section led C. Vafa in 1996 to the idea of F-theory. This is a hy-
pothetical or rather auxiliary twelve dimensional theory which, when compactified on a two-dimensional
torus, gives the Type IIB superstring. The modular group of the torus is identified with the SL(2, ZZ) sym-
metry of Type IIB. However, this twelve-dimensional interpretation is not meant in the sense of a standard
Copyright line will be provided by the publisher
4 Ralph Blumenhagen: Basics of F-theory

7-branes number monodromies

MA = 10 11


A 1
2 1

B 1 MB = −1 0
0 1


C 1 MC = −1 2
An MAn = 10 n1


n
1 1

AB 2 MA MB = −1 0
0 1

A2 B 3 MA2 MB = −1 0
0 1

A2 BA 4 2
MA MB MA = −1 −1
n
MAn MB MC = −1 −n+4


A BC n+2 0 −1

A5 BCB 8 MA5 MB MC MB = −11 0

−1

A6 BCB MA6 MB MC MB = 01 −1


9 0

A6 BCBA MA6 MB MC MB MA = 01 −1


10 1

Table 1 Monodromies around stacks of 7-branes of types A, B,C.

Kaluza-Klein reduction, as first there does not exist a twelve dimensional supergravity theory with sig-
nature (1, 11) in the first place and second in ten dimensions there is no scalar field corresponding to the
volume modulus of this T 2 . Hence the 12-dimensional interpretation serves just to provide a geometriza-
tion of the Type IIB SL(2, ZZ) duality symmetry rather than to correspond to a real compactification from
twelve to ten dimensions. What makes true sense though is to start with the eleven-dimensional M-theory
compactified on T 2 and define F-theory as the vol(T 2 ) → 0 limit.
The true power of this F-theory picture reveals itself when compactifying the Type IIB superstring
to lower dimensions. We have just recalled that there should exist a compactification of Type IIB on
CIP1 with 24 7-branes preserving half the supersymmetry, i.e. 16 supercharges. Observing that M-theory
compactified on a K3 surface breaks half the supersymmetry, one finds that F-theory compactified on an
elliptically fibered K3 with base CIP1 is the Type IIB string compactified on the base CIP1 with 24 7-branes.
For this to make sense, we have to find the 7-branes in this purely geometric description, where we
recall that in Type IIB there exist not just ordinary D7-branes but also these (p, q) 7-branes introduced in
the previous section. We have seen that close to a D7-brane at position u1 ∈CIP1 the complexified dilaton
behaves like j(τ ) ≃ 1/(u − u1). Now τ is really the modular parameter of a geometric elliptic curve and
it is known from mathematics that the j-function naturally appears in this context. For this purpose we
explicitly write the elliptic curve as the hypersurface IP1,2,3 [6] in the homogeneous coordinates (z, x, y).
The fibration over the base B =CIP1 can then be written as the hypersurface constraint
y2 + a1 xyz + a3 yz3 = x3 + a2 x2 z2 + a4 xz4 + a6 z6 , (7)
where the coefficients an are homogeneous polynomials of degree 2n of the two homogeneous coordinates
(u1 , u2 ) on CIP1 . More correctly stated, the an are sections of KB−n , where KB denotes the canonical bundle
of the base B =CIP1 . Note that z = 0 defines a section of the elliptic fibration, i.e. the divisor z = 0 is
the base CIP1 . Completing the square and the cubic term, this so-called Tate form can be written in the
so-called Weierstraß form1
y2 = x3 + f4 xz4 + g6 z6 . (8)
1 For baseCIP1 the Weierstraßform is sufficient, but for compactifications to six and four dimensions the Tate is very convenient.

Copyright line will be provided by the publisher

fdp header will be provided by the publisher 5

To express f4 and g6 in terms of the an , it is convenient to introduce the objects b2 = a21 + 4a2 , b4 =
a1 a3 + 2a4 and b6 = a23 + 4a6 so that

1 1
24 b4 − b22 , 216 b6 − 36 b4b2 + b32 .



f4 = g6 = (9)
48 864
Given the Weierstraß form with sections f4 and g6 , the complex structure τ of the elliptic fiber over a
point (u1 , u2 ) is implicitly given by

4 (24 f4 )3
j(τ ) = , (10)
4 f43 + 27g26

where indeed the j-function appears. Now, the location of the 7-branes should be at the zeros of the
denominator
1
∆ = 4 f43 + 27g26 = − b22 (b2 b6 − b24) − 8b34 − 27b26 + 9b2b4 b6 . (11)
4
This is the so-called discriminant of the elliptic fibration and its zeros are mathematically precisely the
points where the torus degenerates. Note that ∆ is a polynomial of degree 24 in (u1 , u2 ), and thus has 24
zeros. These points are the positions of the 24 7-branes on CIP1 .
So far we assumed that the discriminant has 24 different zeros. However, when some of these zeros
coincide the elliptic fibration further degenerates, i.e. certain 2-cycles shrink to zero size. In the M-theory
description, M2-branes wrapped on these shrunken 2-cycles provide new massless states, which give rise
non-abelian gauge symmetries. In fact there exists a classification by Kodaira [12] of the different types
such an elliptic fibration over CIP1 can degenerate. As shown in table 2, this classification is of the A-D-E
type expected for singularities on K3 respectively enhanced gauge symmetries.
ord( f ) ord(g) ord(∆) fiber singularity comp. local geometry monod.
1 0

≥0 ≥0 0 I0 smooth 1 01
2 2 11


0 0 1 I1 dbl. point 1 y = x +z 01
1 n

0 0 n In An−1 n y2 = x2 + zn 01
1 1

≥1 1 2 II cusp 1 −1 0
0 1

≥1 ≥2 3 III A1 2 y2 = x2 + z2 −1 0
0 1
y2 = x2 + z3


≥2 2 4 IV A2 3 −1 −1
−1 0

2 3 6 I∗0 D4 5 y2 = x2 z + z3 0 −1
2 ≥3 −1 −n

n+6 I∗n Dn+4 n+5 y2 = x2 z + zn+3 0 −1
≥2 3
−1 −1
IV∗ y2 = x3 + z4


≥3 4 8 E6 7 1 0
0 −1
III∗ y2 = x3 + xz3


3 ≥5 9 E7 8 1 0
0 −1

≥4 5 10 II∗ E8 9 y2 = x3 + z5 1 1

Table 2 The Kodaira classification of singular fibers in elliptic surfaces. The local geometry of the elliptic surface
around such an A-D-E singularity is modeled in terms of coordinates (x, y, z) ∈C3 . In the last column the elliptic
monodromy of the singular fiber is given in terms of a SL(2, ZZ)-matrix.

Note that in particular the exceptional gauge groups E6 , E7 and E8 can be realized as enhanced gauge
symmetries in F-theory. Clearly they cannot be realized by fundamental open strings of the perturbative
Copyright line will be provided by the publisher
6 Ralph Blumenhagen: Basics of F-theory

Type IIB string, i.e. not just with (1, 0) strings and D7-branes. These enhancements must involve more
general (p, q) seven branes and the corresponding string-junctions between them.
To get an idea how this works, we compare the geometric monodromy matrices in table 2 with those
listed for stacks of A, B,C branes in table 1. It is evident that for the A-D-E series and the three fiber
types II, III, IV we have a perfect match. Moreover, the number of 7-branes is in all cases identical to the
vanishing order of the discriminant. One can also show [11] that the string junctions ending on the stacks
of branes provide precisely the massless states to fill out the adjoint representation of A-D-E gauge groups.
Note that the Dn series is realized by the An BC 7-branes, which indicates that the BC pair can be considered
as the non-perturbative description of an O7 plane. Therefore, F-theory goes beyond the perturbative Type
IIB orientifolds in that it allows for general (p, q) 7-branes and their corresponding (p, q)-strings. It is
precisely this more general structure which realizes the exceptional gauge groups and as a consequence
all their group theoretic consequences, such as matter in spinor representations of SO(10) or the 10 10 5H
Yukawa coupling for SU(5) GUT models.

4 F-theory compactifications and the Sen limit

Finally, to connect to the second lecture on F-theory, let us briefly comment on lower dimensional com-
pactifications of F-theory. Instead of fibering the torus over a complex one-dimensional base, one can
consider fibrations over surfaces B2 or three-dimensional bases B3 . Supersymmetry then implies that the
total space should either be a Calabi-Yau three-fold (for B2 ) or Calabi-Yau four-fold (for B3 ). One can still
write down a Weierstraß model, where f4 and g6 are sections of KB−4 and KB−6 . The zeros of the discrim-
inant define complex co-dimension one curves in B2 respectively surfaces in B3 and give the location of
7-branes. In these cases, it is more convenient to use the Tate form (7) of the elliptic fibration, as there
exists a refinement of the Kodaira classification, the so-called Tate algorithm, which allows to determine
the gauge group essentially from the vanishing order of the discriminant and the sections an (see [13] for
more details). Now it can however happen that the singularity enhances further where these co-dimension
one objects intersect. Similar to intersecting D-branes, this is where additional matter fields are localized.
In the case of an F-theory compactification on a smooth Calabi-Yau four-fold Y a couple of new issues
need to be considered. First, one can show that chiral matter only arises on the intersection curve between
two 7-branes, if there exists a non-trivial G4 -form background (M-theory point of view). Second, one finds
a non-trivial D3-brane tadpole cancellation condition, which in this case reads
1 χ (Y )
Z
ND3 + G4 ∧ G4 = (12)
2 Y 24
where χ (Y ) denotes the Euler characteristic of the smooth (appropriately resolved) four-fold Y .
One can define a limit in which the string coupling goes to zero almost everywhere on the base. This is
the so-called Sen-limit [14], defined by rescaling a3 → ε a3 , a4 = ε a4 , a6 = ε 2 a6 and sending ε → 0. In
this parameterization one finds
1 1
24 ε b4 − b22 , 216 ε 2 b6 − 36 ε b4 b2 + b32 .



f4 = g6 = (13)
48 864
so that the discriminant becomes
ε2 2 b42
∆=− b (b2 b6 − b24) + O(ε 3 ) ⇒ j(τ ) ≃ . (14)
4 2 ε 2 (b 2
2 b6 − b4 )

Therefore, for ε → 0 the Type IIB string coupling constant gs goes to zero almost everywhere except on
the locus where b2 vanishes. Studying the monodromies one finds a D7-brane on the locus (b2 b6 − b24) = 0
and an O7-plane where b2 = 0. Therefore, the Sen-limit defines the region in the complex structure moduli
space, where F-theory is (almost everywhere) weakly coupled and a perturbative Type IIB orientifold
description is justified.
Copyright line will be provided by the publisher
fdp header will be provided by the publisher 7

Acknowledgements I would like to thank Benjamin Jurke and Dieter Lüst for useful comments about the manuscript.

References
[1] R. Blumenhagen, M. Cvetic, S. Kachru and T. Weigand, Ann. Rev. Nucl. Part. Sci. 59, 269 (2009)
[arXiv:0902.3251 [hep-th]].
[2] R. Donagi and M. Wijnholt, arXiv:0802.2969 [hep-th].
[3] C. Beasley, J. J. Heckman and C. Vafa, JHEP 0901, 058 (2009) [arXiv:0802.3391 [hep-th]].
[4] C. Beasley, J. J. Heckman and C. Vafa, JHEP 0901, 059 (2009) [arXiv:0806.0102 [hep-th]].
[5] R. Donagi and M. Wijnholt, arXiv:0904.1218 [hep-th].
[6] J. J. Heckman, arXiv:1001.0577 [hep-th].
[7] C. Vafa, Nucl. Phys. B 469, 403 (1996) [arXiv:hep-th/9602022].
[8] F. Denef, arXiv:0803.1194 [hep-th].
[9] R. Blumenhagen, B. Körs, D. Lüst and S. Stieberger, Phys. Rept. 445, 1 (2007) [arXiv:hep-th/0610327].
[10] B. R. Greene, A. D. Shapere, C. Vafa and S. T. Yau, Nucl. Phys. B 337, 1 (1990).
[11] M. R. Gaberdiel and B. Zwiebach, Nucl. Phys. B 518, 151 (1998) [arXiv:hep-th/9709013].
[12] K. Kodaira, Annals of Math. 77, 563 (1963).
[13] M. Bershadsky, K. A. Intriligator, S. Kachru, D. R. Morrison, V. Sadov and C. Vafa, Nucl. Phys. B 481, 215
(1996) [arXiv:hep-th/9605200].
[14] A. Sen, Nucl. Phys. B 475, 562 (1996) [arXiv:hep-th/9605150].

Copyright line will be provided by the publisher