On geometry of congruences of null strings in

4-dimensional complex and real pseudo-Riemannian

spaces.
Adam Chudecki∗
September 20, 2021

∗ Center of Mathematics and Physics, Lodz University of Technology,

Al. Politechniki 11, 90-924 Lódź, Poland, [Link]@[Link]
arXiv:1610.02498v2 [gr-qc] 4 Jul 2017

Abstract.
4-dimensional spaces equipped with 2-dimensional (complex holomorphic or real smooth)
completely integrable distributions are considered. The integral manifolds of such distri-
butions are totally null and totally geodesics 2-dimensional surfaces which are called the
null strings. Properties of congruences (foliations) of such 2-surfaces are studied. Some
relations between properties of congruences of null strings, Petrov-Penrose type of SD
Weyl spinor and algebraic types of the traceless Ricci tensor are analyzed.

PACS numbers: [Link], [Link], [Link]

Key words: para-Hermite spaces, para-Kähler spaces, null strings, traceless Ricci tensor.

1 Introduction
The present paper is devoted to some geometrical aspects of complex and real 4-
dimensional spaces. We assume that spaces are equipped with the holomorphic (in com-
plex case) or smooth (in real case) metric. Except the metric tensor we assume the
existence of 2-dimensional, completely integrable distributions. Their integral manifolds
are totally null and totally geodesic 2-dimensional surfaces, the null strings. The family
of such surfaces constitute the congruence (foliation) of the null strings. 2-dimensional
distribution is related to the corresponding 2-form, which is self-dual (SD) or anti-self-
dual (ASD). In this paper we analyze the SD congruences of the null strings. In what
follows we use abbreviation cns (cns := congruence of the SD null strings).
The idea of m-dimensional real spaces equipped with n-dimensional (n < m) totally
null, parallely propagated distributions appeared in the fifties [32] and such spaces are
known nowadays as Walker spaces. For our purposes the most interesting is the case with
m = 4 and n = 2, but in general we do not assume that the distribution is parallel. In
dimension 4, the only real spaces which admit cns are spaces equipped with the metric
of the neutral (split, ultrahyperbolic) signature (+ + −−). Moreover, such structures are
admitted by 4-dimensional complex manifolds.

1
 In complex geometry cns play a great role. These objects appear in the theory of
heavenly spaces (H-spaces) and in the twistor theory as a α(β)-surfaces. The relation
between cns and the algebraic degeneration of SD Weyl spinor in Einstein spaces has been
found in [19]. The further analysis of the spaces which admit cns led to the concept of the
hyperheavenly spaces (HH-spaces). Some properties of cns have been studied in [2,22] but
the significant progress in understanding the geometry of cns has been done in [21, 27].
In [21] the properties of cns are related to the properties of shear-free null geodesic
congruences (abbreviate by sngc). Clearly, cns are complex analogs of sngc. They have
been considered as the most important structures in so-called Plebański programme. The
main idea of the Plebański programme was to find the general techniques of generating
the real solutions of the Einstein field equations from the complex ones. Unfortunately,
such techniques are still unknown despite over 40 years of thorough investigations.
Real 4-dimensional Walker spaces equipped with 2-dimensional integrable and to-
tally null distributions have been recently analyzed intensively (see [3, 6] and references
therein). It appeared, that the most natural formalism in description of the 4-dimensional
Walker spaces is the spinorial formalism. Spinorial formalism as a mathematical tool has
been developed since the sixties [14–16,18]. Spaces with both SD and ASD 2-dimensional
integrable distributions have been introduced in [6] (see also [9]). Recently some 4-
dimensional spaces equipped with the metric of neutral signature and two distinct cns
appeared naturally in the works devoted to geometrical model of bodies which roll on
each other without slipping and twisting [10, 11].
Cns play an important role in Lorentzian geometry as well. The main idea of distin-
guished papers [13,31] was to treat sngc as an intersection of the SD and ASD congruences
of the null strings which exist in the complexification of the tangent bundle. In [31] Robin-
son manifold have been introduced and their relation with cns has been pointed out. Also
the SD null string equations (3.2) have been used to obtain a special class of solutions of
the Rarita-Schwinger equation [29].
Cns play also a great role in the generalizations of the Goldberg - Sachs theorem [8].
For complex, 4-dimensional Einstein spaces the analog of Goldberg - Sachs theorem was
presented in [19]. Then this idea has been generalized to the case of nonzero traceless
Ricci tensor [25] and to the lower [12] and higher dimensions [30]. In [26] the authors
considered ”sharp” versions of the Goldberg - Sachs theorem in dimension 4.
All mentioned above facts prove that cns carry a very important structure which
deserves closer analysis. Our previous work [4] was devoted to Einstein para-Hermite
and Einstein para-Kähler spaces. Such spaces are equipped with two distinct cns (see
also [1] and references therein). In this paper we analyze the non-Einsteinian case. There
are two main goals of the present paper. The first is to analyze the relation between
objects which characterize cns (expansion, Sommers vector) and algebraic types of the
traceless Ricci tensor (in complex case small portion of such analysis has been presented
in [22]). The second aim is to analyze the spaces equipped with three [27] or four distinct
cns.
Our considerations are purely local. The results are valid for the 4-dimensional com-
plex manifolds and can be easily carried over to the case of the real manifold with neutral
signature metric. Usually it is enough to replace all the holomorphic functions by real
smooth functions and complex coordinates by the real ones. Sometimes there appear
subtle differences between complex and real spaces and these differences are pointed out
and thoroughly analyzed.
The paper is organized, as follows.

2
 Section 2 contains basic information on the spinorial formalism. We use spinorial
formalism in Infeld - Van der Waerden - Plebański notation rather then Penrose notation.
In section 3 we consider the spaces which admit one cns. Relation between properties of
the cns and algebraic types of the traceless Ricci tensor are pointed out. Finally a new
Theorem 3.7 concerning the relation between cns and null eigenvectors of the traceless
Ricci tensor is proved. Particularly, we find all algebraic types of the traceless Ricci
tensor generated by the one nonexpanding cns. Subsection devoted to the Weyl spinor is
less original and it contains results known earlier. However, these results are spread all
over the wide literature, so we present them here for completeness.
In section 4 we analyze the spaces equipped with two distinct cns. It appears, that
two distinct cns determine the traceless Ricci tensor completely (4.7) and it is the main
result of this section. All algebraic types of the traceless Ricci tensor generated by two
distinct cns are found. Finally, the relation between null eigenvectors of the traceless
Ricci tensor and properties of two distinct cns is proved.
Section 5 is devoted to spaces which admit a richer structure then the para-Hermite
spaces, i.e. they admit the existence of three or four distinct cns. We find the general
form of the metric which admits two expanding cns and one nonexpanding cns (metric
(5.13)). Finally, we consider the case which - according to our best knowledge - has not
been considered earlier: the space equipped with four distinct cns. In this case we arrive
at the metric (5.21) with constraint equations (5.22). Concluding remarks end the paper.

2 Formalism
In this section we present foundations of the spinorial formalism which seem to be
crucial in geometry based on cns. For more detailed treatment see [16, 18, 20].
Let M be a 4-dimensional complex analytic differentiable manifold endowed with a
holomorphic metric ds2 or a real 4-dimensional smooth differentiable manifold endowed
with a real smooth metric ds2 . Thus one deals with complex relativity (CR) or with
real relativity (RR). More precisely, inside real relativity we distinguish hyperbolic
relativity (HR+ if signature of the metric is (+ + +−) or HR− if the signature is (+ −
−−)), ultrahyperbolic relativity UR (neutral, split, signature (+ + −−)) and Euclidean
relativity (ER+ or ER− if signature is (+ + ++) or (− − −−), respectively).
Denote by Λp (M) the space of all p-forms on M (p = 0, 1, 2, 3, 4). Let (e1 , e2 , e3 , e4 ) ∈
Λ1 (M) be the members of the null tetrad so the metric can be written in the form

ds2 = 2e1 e2 + 2e3 e4 (2.1)

The bases of the Λp (M) are given by (A = 1, 2,Ḃ = 1̇, 2̇)

Λ0 (M) : 1 (2.2)
1 AḂ
Λ (M) : g
1 1
Λ2 (M) : S AB := ∈ṘṠ g AṘ ∧ g B Ṡ , S ȦḂ := ∈RS g RȦ ∧ g S Ḃ
2 2
1 1
Λ3 (M) : ǧ AḂ := g AĊ ∧ S Ċ Ḃ ≡ S AC ∧ g C Ḃ
3 3
1 1 1
Λ4 (M) : vol := − S AB ∧ SAB ≡ S ȦḂ ∧ SȦḂ ≡ ǧ AḂ ∧ gAḂ
12 12 8

3
where
√
 
AḂ e4 e2
(g ) := 2 1 (2.3a)
e −e3
 
AB 2e4 ∧ e2 e1 ∧ e2 + e3 ∧ e4
(S ) := 1 (2.3b)
e ∧ e2 + e3 ∧ e4 2e3 ∧ e1
 
ȦḂ 2e4 ∧ e1 −e1 ∧ e2 + e3 ∧ e4
(S ) := (2.3c)
−e1 ∧ e2 + e3 ∧ e4 2e3 ∧ e2
√
 1 
AḂ e ∧ e2 ∧ e4 e2 ∧ e3 ∧ e4
(ǧ ) := 2 (2.3d)
−e1 ∧ e3 ∧ e4 e1 ∧ e2 ∧ e3
vol := e1 ∧ e2 ∧ e3 ∧ e4 (2.3e)

The spinorial indices are manipulated according to the rules

ΨA =∈AB ΨB ΨȦ =∈ȦḂ ΨḂ , ΨA =∈BA ΨB , ΨȦ =∈Ḃ Ȧ ΨḂ

, (2.4)
   
0 1 AB 0 1
(∈AB ) := =: (∈ ) , (∈ȦḂ ) := =: (∈ȦḂ )
−1 0 −1 0
Nevertheless, one has to carefully raise and lower spinorial indices in the objects from
tangent space. Indeed, if ∂ Ȧ means ∂/∂ΨȦ and ΨȦ =∈ȦḂ ΨḂ then consistency with (2.4)
implies
∂ Ȧ = ∂Ḃ ∈ȦḂ , ∂Ȧ =∈Ḃ Ȧ ∂ Ḃ , ∂ A = ∂B ∈AB , ∂A =∈BA ∂ B (2.5)
The 1-forms g AḂ are holomorphic in CR. In the case of RR the 1-forms g AḂ have the
following properties under complex conjugation:

HR+ : g AḂ = g B Ȧ (2.6)

HR− : g AḂ = −g B Ȧ
UR : g AḂ = g AḂ
ER+ : g AḂ = −gAḂ
ER− : g AḂ = gAḂ

where the overbar stands for the complex conjugation.

Definition 2.1. Hodge star ∗ is the linear map ∗ : Λp (M) → Λ4−p (M) defined as follows
1 h iπ i
∗ω = exp (p(4 − p) − 2) ∈a1 ...ap b1 ...b4−p ω a1 ...ap eb1 ∧ ... ∧ eb4−p (2.7)
p! (4 − p)! 2
where ω ∈ Λ(M) := ⊕4p=0 Λp (M) is the p-form and it has the local representation
1
ω= ωa ...a ea1 ∧ ... ∧ eap (2.8)
p! 1 p
Operation ∗ defined by (2.7) is the idempotent operation on Λ(M): ∗ ∗ ω = ω. Under
(2.7) the p-forms defined by (2.2) behave as follows

∗1 = vol , ∗g AḂ = ǧ AḂ , ∗S AB = S AB , ∗S ȦḂ = −S ȦḂ (2.9)

AḂ AḂ
∗ǧ =g , ∗vol = 1

4
 Some identities involving (2.2) are helpful in calculations

g AḂ ∧ g C Ḋ =∈AC S Ḃ Ḋ + ∈Ḃ Ḋ S AC (2.10)

AḂ Ċ Ḋ Ḃ Ḋ AĊ Ḃ Ċ AḊ AB C Ḋ AC B Ḋ BC AḊ
g ∧S = − ∈ ǧ − ∈ ǧ , S ∧g = − ∈ ǧ − ∈ ǧ
1 AB A B 1 ȦḂ
S ∧ SCD = −δ(C δD) vol , S ∧ SĊ Ḋ = δ(ȦĊ δḊ)
Ḃ
vol , S AB ∧ SĊ Ḋ = 0
4 4
The connection 1-forms Γab in null tetrad formalism are connected with the spinorial
connection 1-forms ΓAB and ΓȦḂ by the relations
   
1 2 Γ42 Γ12 + Γ34 1 2 Γ41 −Γ12 + Γ34
(ΓAB ) = − , (ΓȦḂ ) = −
2 Γ12 + Γ34 2 Γ31 2 −Γ12 + Γ34 2 Γ32
(2.11)
The connection 1-forms can be decomposed according to
1 1
ΓAB = − ΓABM Ṅ g M Ṅ , ΓȦḂ = − ΓȦḂM Ṅ g M Ṅ (2.12)
2 2
The general formula for the spinorial covariant derivative of the arbitrary spinor field
ΨA Ḃ
C Ḋ
reads

∇M Ṅ ΨA Ḃ
C Ḋ
AḂ
= ∂M Ṅ ΨC Ḋ
+ ΓASM Ṅ ΨC
S Ḃ
Ḋ
− ΓSCM Ṅ ΨA Ḃ
S Ḋ
(2.13)
+ΓḂṠM Ṅ ΨA Ṡ
C Ḋ
− ΓṠḊM Ṅ ΨA Ḃ
C Ṡ

where
∇AḂ := g aAḂ ∇a , ∂AḂ := g aAḂ ∂a (2.14)
and the matrices gaAḂ are defined by the relation g AḂ = gaAḂ ea . Note that we have

gaAḂ g bAḂ = −2δab , gaAḂ g aC Ḋ = −2δAC δḂḊ (2.15)

The arbitrary vector V has the form

1 1
V = V a ∂a = − V AḂ ∂AḂ , V a = − g aAḂ V AḂ ⇐⇒ V AḂ = gaAḂ V a (2.16)
2 2
The first and second Cartan structure equations read

Dg AḂ = dg AḂ + ΓAC ∧ g C Ḃ + ΓḂĊ ∧ g AĊ = 0 (2.17a)

RAB = dΓAB + ΓAC ∧Γ C
B (2.17b)
RȦḂ = dΓȦḂ + ΓȦĊ ∧ ΓĊḂ

RAB and RȦḂ are the components of curvature 2-forms of the connection ΓAB or ΓȦḂ ,
respectively, D := − 21 g AḂ ∇AḂ and d := − 12 g AḂ ∂AḂ . The 2-forms RAB = R(AB) and
RȦḂ = R(ȦḂ) can be decomposed with respect to the bases S AB and S ȦḂ as follows

1 R 1
RAB = − CABCD S CD + SAB + CAB Ċ Ḋ S Ċ Ḋ (2.18)
2 24 2
1 R 1
RȦḂ = − CȦḂ Ċ Ḋ S Ċ Ḋ + S + CCDȦḂ S CD
2 24 ȦḂ 2

5
The objects CABCD = C(ABCD) , CȦḂ Ċ Ḋ = C(ȦḂ Ċ Ḋ) , CAB Ċ Ḋ = C(AB)Ċ Ḋ = CAB(Ċ Ḋ) and
R have a transparent geometrical meaning. Namely, CABCD is the spinorial image of the
self-dual (SD) part of the Weyl tensor, CȦḂ Ċ Ḋ is the spinorial image of the anti-self-dual
(ASD) part of the Weyl tensor, CAB Ċ Ḋ is the spinorial image of the traceless Ricci tensor
and, finally, R is the curvature scalar.
Algebraic classification of totally symmetric 4-index spinors (like CABCD and CȦḂ Ċ Ḋ )
has been presented in [14]. There are 6 different Petrov-Penrose types of such spinors
in CR (namely [I], [II], [D], [III], [N] and [−]). However, in UR there are 10 different
types. These types are [Ir ], [Irc ], [Ic ], [IIr ], [IIrc ], [Dr ], [Dc ], [IIIr ], [Nr ] and [−]. In [26] the
authors used the following symbols for these types: Gr , SG, G, IIr , II, Dr , D, IIIr , Nr
and 0, respectively. We do not present here the details of this classification. They can be
found in [5, 26].
Note, that under the complex conjugation we have
CR : ΓAB , ΓȦḂ , CABCD , CȦḂ Ċ Ḋ , CAB Ċ Ḋ and R are holomorphic
HR : ΓȦḂ = ΓAB , CȦḂ Ċ Ḋ = CABCD , CAB Ċ Ḋ = CCDȦḂ , R = R
UR : ΓAB , ΓȦḂ , CABCD , CȦḂ Ċ Ḋ , CAB Ċ Ḋ and R are real
ER : ΓAB = ΓAB , ΓȦḂ = ΓȦḂ , CABCD = C ABCD , CȦḂ Ċ Ḋ = C ȦḂ Ċ Ḋ
CAB Ċ Ḋ = C AB Ċ Ḋ , R = R
The Ricci identities for 1-index spinors read
1 E
∇ (Ċ ∇|E|Ḋ) ΨA = ΨE C AE Ċ Ḋ (2.19)
2
1 R
∇ Ė ∇D)Ė ΨA = ΨE − C AECD + ∈E(C δ AD)


2 (C 12
1
∇ Ė ∇D)Ė ΨȦ = ΨĖ CCD ȦĖ
2 (C
1 E R
∇ (Ċ ∇|E|Ḋ) ΨȦ = ΨĖ − C ȦĖ Ċ Ḋ + ∈Ė(Ċ δ ȦḊ)


2 12
From (2.19) we can easily obtain the Ricci identities for any spinor ΨA... Ḃ...
C...Ḋ...
.
Let mA and µA be the pair of linearly independent spinors such that mA µA = 1. SD
curvature coefficients C (i) , i = 1, ..., 5 are defined as follows
2CABCD =: C (1) µA µB µC µD + 4C (2) m(A µB µC µD) + 6C (3) m(A mB µC µD) (2.20)
+4C (4) m(A mB mC µD) + C (5) mA mB mC mD
The relation between the traceless Ricci tensor Cab and its spinorial image CAB Ċ Ḋ
reads
1
Cab = gaAĊ gb B Ḋ CAB Ċ Ḋ ⇐⇒ CAB Ċ Ḋ = Cab g aAĊ g bB Ḋ (2.21)
4
We decompose the traceless Ricci tensor CAB Ċ Ḋ according to
CAB Ṁ Ṅ = mA mB AṀ Ṅ + 2m(A µB) BṀ Ṅ + µA µB CṀ Ṅ (2.22)
where AṀ Ṅ , BṀ Ṅ and CṀ Ṅ are symmetric, 2-index dotted spinors (see [16] for more
details). From those spinors the following scalars can be constructed

a := AȦḂ AȦḂ , b := BȦḂ B ȦḂ , c := CȦḂ C ȦḂ (2.23)

ȦḂ ȦḂ ȦḂ
r := AȦḂ B , n := AȦḂ C , s := BȦḂ C

6
The equations for the eigenvectors and eigenvalues of the traceless Ricci tensor have the
form (compare (2.15) and (2.16))
1
C ab V b = λ V a ⇐⇒ CAB Ċ Ḋ V B Ḋ = − λ VAĊ (2.24)
2
The characteristic polynomial of the matrix (C ab ) of the traceless Ricci tensor reads
4
X
W(x) := det(C ab − xδ ab ) = (−1)i C x4−i ≡ Cx4 − Cx3 + Cx2 − Cx + C (2.25)
[i] [0] [1] [2] [3] [4]
i=0

where coefficients C read

[i]

C := 1 , C := C a1[a1 ...C akak ] , k = 1, 2, 3, 4 (2.26)

[0] [k]

After long but elementary calculations we find the relations between C and the scalars
[i]
given by (2.23)

C=1, C=0, C = 4b − 4n , C = −16rs + 8nb + 4ac + 4b2 (2.27)

[0] [1] [2] [4]
1 2 1 1 1 1
C = 16AṀ Ṅ C ṄṘ B ṘṀ =⇒ 2
C = rsn + abc − cr2 − bn2 − as2
[3] 16 [3] 2 2 2 2

Definitions of undotted and dotted Plebański spinors are [16, 17]

VABCD := 4 C(AB Ṁ Ṅ CAC)Ṁ Ṅ = 4a mA mB mC mD + 16r m(A mB mC µD) (2.28a)

+(8n + 16b) m(A mB µC µD) + 16s m(A µB µC µD) + 4c µA µB µC µD
VȦḂ Ċ Ḋ := 4 CM N (ȦḂ C M NĊ Ḋ) = 8A(ȦḂ CĊ Ḋ) − 8B(ȦḂ BĊ Ḋ) (2.28b)

Plebański spinors are 4-index and totally symmetric, so they can be classified analogously
like SD and ASD Weyl spinors.
The algebraic classification of the traceless Ricci tensor plays an important role in ana-
lysis presented in this paper. Such classification in 4-dimensional spaces with Lorentzian
metric has been done in [16, 17] (15 types) and then it was generalized on the complex
case in [24] (17 types). There are 33 different types of the traceless Ricci tensor in 4-
dimensional real spaces equipped with the metric of signature (+ + −−). The detailed
discussion which leads to this classification can be found in [5]. Here we present only brief
summary of the results of [5]. The information about the algebraic type of the traceless
Ricci tensor in UR is gathered in the following symbol
[Aj ]⊗[Bk ]
[n1 E1 − n2 E2 − ...]v(q1 q2 ...)

Inside the square bracket all different eigenvalues Ei , i = 1, 2, ..., N0 of the polynomial
W(x) together with their multiplicities ni are listed. Of course

n1 + n2 + ... + nN0 = 4
n1 E1 + n2 E2 + ... + nN0 EN0 = 0

The last equality follows from the fact, that the matrix (C ab ) is traceless. Complex
eigenvalues are denoted by Z and the real ones by R. Real eigenvalues have additional

7
superscript which denotes the type of the corresponding eigenvector. Rs means, that
the eigenvector which corresponds to the eigenvalue R is space-like, Rt - time-like, Rn
- null, Rns - null or space-like, Rnt - null or time-like and finally Rnst means, that the
eigenvector can be of the arbitrary type. [We use the same definition of space-like and
time-like vectors, like in the Lorentzian case, i.e. V a Va > 0 means, that V a is space-like,
V a Va < 0 stands for time-like vectors and finally, V a Va = 0 means, that the vector is
null]. Superscript v denotes the number of eigenvectors and the numbers qi in round
bracket describes the form of the minimal polynomial. Namely, the minimal polynomial
of the matrix (C ab ) has the form
N0
Y
Wmin (x) := (x − Ei )qi (2.29)
i=1

Finally, the symbol [Aj ] ⊗ [Bk ] describes Petrov-Penrose types of the Plebański spinors,
VABCD and VȦḂ Ċ Ḋ , respectively. For example, [IIIr ] ⊗ [Nr ] means, that VABCD is of the
type [IIIr ] while VȦḂ Ċ Ḋ is of the type [Nr ].

3 Spaces equipped with one congruence of the SD

null strings
3.1 Definition and basic properties
Definition 3.1. Congruence (foliation) of null strings in a complex (real) 4-dimensional
manifold M is a family of totally null and totally geodesics 2-dimensional holomorphic
(smooth) surfaces, such that for every point p ∈ M there exists only one surface of this
family such that p belongs to this surface.

Theorem 3.2 (Plebański, Rózga, [21]). The complex (real) 4-dimensional manifold M
admits a congruence of the SD null strings, if there exists a holomorphic (smooth) 2-form
Σ and a 1-form σ such that

Σ∧Σ=0 (3.1a)
∗Σ = Σ (3.1b)
dΣ = σ ∧ Σ (3.1c)

In what follows we often abbreviate the cns defined by the 2-form Σ by Σ-congruence.
The 2-form Σ which satisfies the conditions (3.1a) and (3.1b) from the Theorem 3.2
has the form Σ = mA mB S AB where mA is nowhere vanishing spinor. The condition
(3.1c) implies, that 2-form Σ is an element of the 2-surface completely integrable in the
Frobenius sense. It gives restrictions on the spinor mA . Indeed, one finds that spinor mA
has to satisfy the following equations

mA mB ∇AṀ mB = 0 (3.2)

The crucial equations (3.2) are called the SD null string equations. From (3.2) we find

∇AṀ mB = ZAṀ mB + ∈AB MṀ (3.3)

8
where ZAṀ is the Sommers vector [21] and the spinor MṀ is the expansion of the con-
gruence of the SD null strings (see [21]). With fixed Riemannian structure the expansion
describes the most important and invariant property of the cns. If MṀ = 0 then the
2-dimensional distribution DmA := {mA aḂ , mA bḂ }, aḂ bḂ 6= 0 is parallely propagated. It
means, that ∇X V ∈ DmA for every vector field V ∈ DmA and for arbitrary vector field
X. Such cns are called nonexpanding or plane. If MṀ 6= 0 then we deal with expanding
(or deviating) cns. Under re-scaling mA = mm0A the expansion and Sommers vector
transform as follows
1
MȦ0 = M , ZA0 Ḃ = ZAḂ − ∇AḂ ln m (3.4)
m Ȧ

3.2 SD Weyl spinor

Consider the integrability conditions of the SD null strings equations. Acting on (3.3)
with ∇M Ṅ and using Ricci identities (2.19), one arrives at the equations
R
−2mS CCSM A + m(M ∈A)C = mC ∇(MṄ ZA)Ṅ − ∈C(M (ZA)Ṅ M Ṅ − ∇A)Ṅ M Ṅ ) (3.5a)
6
2mS CCS Ṅ Ḃ = mC ∇A(Ṅ Z|A|Ḃ) + ZC(Ḃ MṄ ) − ∇C(Ḃ MṄ ) (3.5b)

Immediately we find
Theorem 3.3. If a spinor mA generates a congruence of SD null strings, then it is a
Penrose spinor.
Proof. Contracting the integrability conditions of the SD null strings equations (3.5a)
with mC mM mA we find mS mC mM mA CCSM A = 0. 
Theorem 3.4. If a spinor mA generates a nonexpanding congruence of SD null strings
then it is a multiple Penrose spinor.
Proof. Contracting the integrability conditions of the nonexpanding (M Ḃ = 0) SD null
strings equations (3.5a) with mC mM we find mS mC mM CCSM A = 0. 
Define spinor µA by the relation µA mA = 1. Then spinors mA and µA constitute the
basis of the 1-index undotted spinors. The SD curvature coefficients C (i) , i = 1, ..., 5 are
defined as usual by (2.20). Eqs. (3.5a) constitute the set of 6 equations which relate SD
Weyl spinor and the curvature scalar with the Sommers vector and the expansion of the
cns. We easily conclude that

C (1) = 0 , ∇(MṄ ZA)Ṅ = C (4) mA mM + 3C (3) m(A µM ) + 3C (2) µA µM (3.6)

Note that the coefficient C (5) remains undetermined, so the existence of only one cns does
not determine SD conformal curvature. Using (3.6), the integrability conditions (3.5a)
reduce to the equations
1
ZAḂ M Ḃ − ∇AḂ M Ḃ = (R − 6C (3) )mA − 2C (2) µA (3.7)
6
From (3.7) (with help of (3.6)) we find the curvature scalar
3
R = µA (3∇AṄ − Z AṄ )(2mM ZM Ṅ − 3MṄ ) + 2ZN Ṅ Z N Ṅ + 3∇N Ṅ Z N Ṅ (3.8)
2
9
Finally (3.7) leaves us with one constraint equation

∇M Ṅ mM (2mA ZAṄ − 3MṄ ) = 0

(3.9)

Gathering: from 6 equations (3.5a) we get the formulas (3.6) for curvature coefficients
C (1) , C (2) , C (3) and C (4) , the curvature scalar R given by (3.8) and one constraint equation
(3.9).
Petrov-Penrose types of CABCD of the spaces equipped with one cns are gathered in
the Table 1. In the case of the Einstein spaces the types [I] (in CR), [Ir ] and [Irc ] (in
UR) are not admitted (compare generalized Goldberg - Sachs Theorem which has been
formulated for the first time in [19]).

Types in CR
Curvature scalar MȦ 6= 0 MȦ = 0
R 6= 0 [I], [II], [D], [II], [D]
R=0 [III], [N], [−] [III], [N], [−]
Types in UR
Curvature scalar MȦ 6= 0 MȦ = 0
R 6= 0 [Ir ], [Irc ], [IIr ], [IIrc ], [IIr ], [IIrc ], [Dr ]
R=0 [Dr ], [IIIr ], [Nr ], [−] [IIIr ], [Nr ], [−]

Table 1: Petrov-Penrose types of the SD Weyl spinor of the spaces equipped with one
congruence of SD null strings.

3.3 Traceless Ricci tensor

Now we find the possible types of the traceless Ricci tensor which are admitted by
the space equipped with one cns in CR and UR. The criteria described in [24] allow to
establish the algebraic type of the traceless Ricci tensor in CR. In [5] we presented the
similar criteria for the Plebański-Przanowski types of the traceless Ricci tensor in UR.
If a space is equipped with one cns then the traceless Ricci tensor is determined with
precision up to the factor mA CAB Ṁ Ṅ and it is given by (3.5b). Using the decomposition
(2.22) we find BȦḂ and CȦḂ as

2BȦḂ = µN (Z N(Ȧ MḂ) − ∇N(Ȧ MḂ) ) + ∇N (Ȧ Z NḂ) (3.10a)

2CȦḂ = mN (∇N(Ȧ MḂ) − Z N(Ȧ MḂ) ) (3.10b)

Spinor AȦḂ is not determined and, consequently, the spaces equipped with one cns only
admit all Plebański-Przanowski types of the traceless Ricci tensor. Stronger restrictions
on the traceless Ricci tensor appear if spaces admit the nonexpanding cns. Indeed, one
finds that MȦ = 0 =⇒ CȦḂ = 0 =⇒ n = c = s = 0 what implies that the only nonzero
scalars C are
[i]
C = 4b , C = 4b2 (3.11)
[2] [4]

10
The characteristic polynomial (2.25) takes the form
W(x) = x4 + 4b x2 + 4b2 = (x2 + 2b)2 (3.12)
The Plebański spinors read


VABCD = 4m(A mB a mC mD) + 4r mC µD) + 4b µC µD) (3.13a)
VȦḂ Ċ Ḋ = −8B(ȦḂ BĊ Ḋ) (3.13b)
Using criteria given in [5] and the form of the Plebański spinors we find that UR spaces
equipped with one nonexpanding cns admit only 12 types of the traceless Ricci tensor
(see Table 2). In the case of the CR there are only 9 different types of the traceless Ricci
tensor.
There is an interesting connection between the existence of nonexpanding cns and the
eigenvalues of the characteristic polynomial of the traceless Ricci tensor. Indeed, we have
Theorem 3.5 (Przanowski, [22]). In CR the existence of nonexpanding congruence of
SD null strings implies that the characteristic polynomial
√ of the traceless Ricci tensor has
two double or one quadruple eigenvalue, namely λ = ± −2b. 
Corollary 3.6 (Przanowski, [22]). In CR the existence of nonexpanding congruence of
SD null strings implies that the traceless Ricci tensor has at least one null eigenvector
which is tangent to the null strings. Moreover:
• if there is exactly one null eigenvector then it is tangent to the null string
• if there are two linearly independent null eigenvectors then at least one of them is
tangent to the null string
• if there are three or four linearly independent null eigenvectors then exactly two of
them are tangent to the null string 
Remark. In UR Theorem 3.5 still holds true but Corollary 3.6 does not hold
true anymore. Indeed, there are two algebraic types (namely [Dr ]⊗[Dc ] [2Z − 2Z̄]4(11) and
[Dc ]⊗[Dc ]
[2R1s − 2R2t ]4(11) ) which are generated by the nonexpanding congruence of SD null
strings, but they do not admit real null eigenvector of C ab . However, the following the-
orem holds true in both CR and UR (slightly weaker version of this theorem has been
presented in [22]).
Theorem 3.7. Let µA rḂ be a null eigenvector of the traceless Ricci tensor and let mA
be any spinor such that mA µA 6= 0. Then mA rḂ is also a null eigenvector of the traceless
Ricci tensor iff CȦḂ rḂ = 0 where CȦḂ is defined by mA and µA according to (2.22).
Proof. Without any loss of generality we put µA mA = 1. Because the vector µA rȦ is null
eigenvector of the traceless Ricci tensor then
1
(mA mB AȦḂ + 2m(A µB) BȦḂ + µA µB CȦḂ )µA rȦ = − λµB rḂ (3.14)
2
1 √
=⇒ AȦḂ rȦ = 0 , BȦḂ rȦ = − λrḂ , λ = ± −2b
2
From (3.14) if follows that
1
CAB ȦḂ mA rȦ = λmB rḂ ⇐⇒ CȦḂ rḂ = 0 (3.15)
2


11
 Remark. Obviously, Theorem 3.7 holds true for the CȦḂ = 0, i.e., for the non-
expanding cns and in this form it was presented in [22]. However, for the proof it is
enough to assume CȦḂ rḂ = 0 =⇒ c = 0. It means, that Theorem 3.7 holds true for
special class of expanding cns such that c = 0.
In Table 2 we gather information about the number of null eigenvectors of the traceless
Ricci tensor in UR and CR. 2T T means, that both null eigenvectors are tangent to the
null string and 2T N means, that only one of them is tangent to the null string. In CR
there is subtle difference between types (2) [4N ]a2 and (2) [4N ]b2 and this difference has not
been recognized earlier in [22].

Criteria Types in UR Types in CR

b 6= 0 ab 6= r2 [IIr ]⊗[Dr ]
[2R1n − 2R2n ]2(22) 2T T [2N1 − 2N ]b4 2T T
ab = r2 , b > 0 [Dr ]⊗[Dc ]
[2Z − 2Z̄]4(11) 0 [2N1 − 2N ]2 4
ab = r2 , b < 0 [Dc ]⊗[Dc ]
[2R1s − 2R2t ]4(11) 0 [2N1 − 2N ]2 4
[Dr ]⊗[Dr ]
[2R1nst − 2R2nst ]4(11) 4
[Dr ]⊗[Dr ]
[2R1nst − 2R2n ]3(12) 3 [2N1 − 2N ](1−2) 3
[IIIr ]⊗[Nr ]
b = 0 BȦḂ 6= 0 r 6= 0 [4Rn ]1(4) 1 [4N ]b4 1
[Nr ]⊗[Nr ]
r = 0, a 6= 0 [4Rnt ]2(3) 1 [4N ]3 1
[Nr ]⊗[Nr ]
[4Rns ]2(3) 1
[−]⊗[Nr ]
r=a=0 [4Rn ]2(2) 2T N (2)
[4N ]a2 2T N
[Nr ]⊗[−]
BȦḂ = 0 r = 0, a 6= 0 [4Rn ]2(2) 2T T (2)
[4N ]b2 2T T
[−]⊗[−]
r=a=0 [4Rnst ]4(1) 4 [4N ]1 4
[−]⊗[−]
[4Rnst ]3(2) 3 (3)
[4N ]2 3

Table 2: Possible types of the traceless Ricci tensor in spaces equipped with one non-
expanding congruence of the SD null strings.

4 Spaces equipped with two congruences of the SD

null strings
4.1 Basic concepts and integrability conditions
Definition 4.1. Two congruences of SD null strings Σ and Σ
e are distinct (complemen-
tary, transversal), if Σ ∧ Σ
e 6= 0.

Theorem 4.2. Let Σ-congruence be generated by the spinor mA and the Σ-congruence
e
A
be generated by the spinor µA . These congruences are distinct, iff mA µ 6= 0.

Proof. Because both congruences are self-dual, then Σ = mA mB S AB and Σ

e = µA µB S AB .
e = 2(mA µA )2 vol. Consequently Σ ∧ Σ
Using (2.10) we find Σ ∧ Σ e 6= 0 =⇒ mA µA 6= 0.

12
Conversely, consider two spinors mA and µA which generate cns such that mA µA 6= 0 .
Then mA µA 6= 0 =⇒ Σ ∧ Σ e 6= 0. 
Spaces which admit two distinct congruences of the SD (or ASD) null strings are
known as a (complex or real) para-Hermite spaces [7, 23]. Spinors mA and µA can be
normalized mA µA = 1 without any loss of generality. We deal now with the equations
for the Σ-congruence and Σ-congruence
e

∇AḂ mC = ZAḂ mC + ∈AC MḂ (4.1a)

∇ µC = Ze µC + ∈AC M
AḂ AḂ
f
Ḃ (4.1b)

and with the integrability conditions of the equations (4.1a) and (4.1b)

R
−2mS CCSM A + m(M ∈A)C = mC ∇(MṄ ZA)Ṅ − ∈C(M (ZA)Ṅ M Ṅ − ∇A)Ṅ M Ṅ ) (4.2a)
6
R
−2µS CCSM A + µ(M ∈A)C = µC ∇(MṄ ZeA)Ṅ − ∈C(M (ZeA)Ṅ M fṄ − ∇ M
A)Ṅ
fṄ ) (4.2b)
6
2mS CCS Ṅ Ḃ = mC ∇A(Ṅ Z|A|Ḃ) + ZC(Ḃ MṄ ) − ∇C(Ḃ MṄ ) (4.2c)
2µS CCS Ṅ Ḃ = µC ∇A(Ṅ Ze|A|Ḃ) + ZeC(Ḃ M
f −∇ M
Ṅ ) C(Ḃ
f
Ṅ ) (4.2d)

Because mA µA = 1 =⇒ ∇B Ḃ (mA µA ) = 0 we find the relation between the expansions

and Sommers vectors for the Σ- and Σ-congruences
e

f − µA M
ZeAḂ + ZAḂ = mA M (4.3)
Ḃ Ḃ

In what follows we use the Eq. (4.3) to eliminate ZeAḂ .

Integrability conditions (4.2a)-(4.2b) constitute the set of 12 equations. One quickly
finds that all curvature coefficients C (i) are determined by

C (1) = C (5) = 0 , ∇(MṄ ZA)Ṅ = C (4) mA mM + 3C (3) m(A µM ) + 3C (2) µA µM (4.4)

Eqs. (4.2a) and (4.2b) reduces to the following ones

6C (3) − R
−2C (2) µA − mA = ZAḂ M Ḃ − ∇AḂ M Ḃ (4.5a)
6
6C (3) − R
2C (4) mA + fḂ + ∇ M
µA = ZAḂ M AḂ
fḂ + µA M M
Ḃ
fḂ (4.5b)
6
Eqs. (4.5a) and (4.5b) have to be consistent with each other and with (4.4). After simple
calculations we find that the curvature scalar R has the form (3.8). We are left with three
constraint equations which can be rearranged into the form

∇M Ṅ mM (2mA ZAṄ − 3MṄ ) = 0

(4.6a)
 
∇M Ṅ µM (2µA ZAṄ + Mf ) =0
Ṅ (4.6b)

∇M Ṅ (mM M
f + µM M ) = 0
Ṅ Ṅ (4.6c)

Gathering, from 12 equations (4.2a) and (4.2b) we obtained SD conformal curvature coef-
ficients C (i) , i = 1, ..., 5 (5 equations), curvature scalar R (1 equation) and the constraints
(4.6) (3 equations). The remaining 3 equations are identically satisfied.

13
 Integrability conditions (4.2c)-(4.2d) give 12 equations for the 9 components of the
traceless Ricci tensor. Three of these equations are identically satisfied and the rest bring
us to the formulas

2AȦḂ := µN ∇N(Ȧ M
f + µN Z N M
Ḃ)
f +M
(Ȧ Ḃ)
f M f
(Ȧ Ḃ) (4.7a)
2CȦḂ := mN ∇N(Ȧ MḂ) − mN Z N(Ȧ MḂ) (4.7b)
2BȦḂ := µN Z N(Ȧ MḂ) − µN ∇N(Ȧ MḂ) + ∇N (Ȧ Z NḂ) (4.7c)

Formulas (4.7) prove important result. Expansions of congruences, the Sommers vector
(MȦ , M
f , Z ) and their covariant derivatives determine completely the form of the
Ḃ AḂ
traceless Ricci tensor.

4.2 SD Weyl spinor

Theorem 4.3. If complex (real) space admits two distinct congruences of SD null strings
then SD Weyl spinor CABCD cannot be of the type [N] ([Nr ]).

Proof. Assume, that SD Weyl spinor is of the type [N] ([Nr ]). Then there exists a nonzero
complex (real) spinor αA such that CABCD αA = 0. Decomposing spinor αA according to
αA = αmA + βµA , putting this into (2.20) and remembering that C (1) = C (5) = 0 we find
αC (i) = 0 = βC (i) , i = 2, 3, 4. All C (i) cannot be simultaneously equal zero, so the only
solution is α = β = 0 what contradicts the assumption that spinor αA in nonzero. 
Because C (1) = C (5) = 0 one arrives to the following formula for SD Weyl spinor

CABCD =: m(A µB 2C (2) µC µD) + 3C (3) mC µD) + 2C (4) mC mD)

(4.8)

All possible types and forms of CABCD are gathered in the Table 3 in which the following
abbreviation has been used

δ := 9C (3) C (3) − 16C (2) C (4) (4.9)

Petrov-Penrose types of CABCD in para-Hermite spaces depends strongly on expan-

sions of the congruences (compare formulas (4.5)). For example, if both congruences
are nonexpanding, then C (2) = 0 = C (4) and 6C (3) = R. Consequently, 2CABCD =
R m(A mB µC µD) so for R 6= 0 we have type [D] in CR or type [Dr ] in UR and for R = 0
we have type [−]. In the case of Einstein spaces, the conditions Cab = 0 and R = −4Λ
reduce the possible Petrov-Penrose types of CABCD . In this case analysis is less straight-
forward (it is necessary to use the Bianchi identities) and we do not present the details
here. Some portion of this analysis can be found in [23]. The results are gather in Tables
4 and 5.

4.3 Traceless Ricci tensor

The traceless Ricci tensor in para-Hermite spaces is completely determined via Som-
mers vector, expansions of both congruences and covariant derivatives of these objects
(compare Eqs. (4.7)). The most general case is when both congruences are expanding.
It seems (somehow surprisingly) that in this case all Plebański-Przanowski types of the
traceless Ricci tensor are admitted. The case with one nonexpanding congruence has been
discussed in subsection 3.3 and this discussion remains valid in spaces equipped with one

14
 Types Conditions SD Weyl spinor
CR UR
−
[I] [Ir ] C (2) 6= 0, C (4) 6= 0, δ > 0 CABCD = 2C (2) m(A µB p+ C pD)
√

[Irc ] C (2) 6= 0, C (4) 6= 0, δ < 0 p±
A := µ A + 4C
1
(2) 3C (3)
± δ mA
[II] [IIr ] C (2) = 0, C (3) 6= 0, C (4) 6= 0 CABCD = m(A mB µC pD)
pA := 3C (3) µA + 2C (4) mA
C (4) = 0, C (2) 6= 0, C (3) 6= 0 CABCD = µ(A µB mC pD)
pA := 2C (2) µA + 3C (3) mA
C (2) 6= 0, C (4) 6= 0, δ = 0 CABCD = 2C (2) m(A µB pC pD)
3C (3)
pA := µA + m
4C (2) A

[D] [Dr ] C (2) = C (4) = 0, C (3) 6= 0 CABCD = 3C (3) m(A mB µC µD)

[III] [IIIr ] C (2) = C (3) = 0, C (4) 6= 0 CABCD = 2C (4) m(A mB mC µD)
C (3) = C (4) = 0, C (2) 6= 0 CABCD = 2C (2) m(A µB µC µD)
[−] [−] C (2) = C (3) = C (4) = 0 CABCD = 0

Table 3: Petrov-Penrose types of the SD Weyl spinor in the para-Hermite spaces.

nonexpanding and one expanding cns. The last case concerns both congruences being
nonexpanding. One finds that MȦ = Mf = 0 implies A = C = 0 and consequently
Ȧ ȦḂ ȦḂ
a = c = n = r = s = 0. Relations (3.11) and (3.12) are still valid but Plebański spinors
take the form

VABCD = 16b m(A mB µC µD) , VȦḂ Ċ Ḋ = −8B(ȦḂ BĊ Ḋ) (4.10)

However, these are not the only restrictions on the traceless Ricci tensor implied by the
existence of two nonexpanding cns. We have transparent corollary which follows from
the Theorem 3.7

Corollary 4.4. Assume, that the traceless Ricci tensor generated by two distinct, non-
expanding congruences of SD null strings admits a null eigenvector. Then

• there are two linearly independent null eigenvectors, the first is tangent to the Σ-
congruence, the second to the Σ-congruence
e

• there are four linearly independent null eigenvectors, the first pair is tangent to the
Σ-congruence, the second pair to the Σ-congruence
e 

Corollary 4.4 proves, that the cases with one null eigenvector or three linearly in-
dependent null eigenvectors of the traceless Ricci tensor are not admitted in the spaces
equipped with two nonexpanding cns.
Using algebraic criteria given [5], the form of Plebański spinors and Corollary 4.4 we
arrive at the Table 6.

15
 Types in CR
Curvature scalar MȦ 6= 0, MȦ 6
f = 0 MȦ = 0, MȦ 6
f = 0 MȦ = 0 = M
f
Ȧ

R 6= 0 [I], [II], [D], [II], [D] [D]

R=0 [III], [−] [III], [−] [−]
Types in UR
Curvature scalar MȦ 6= 0, MȦ 6
f = 0 MȦ = 0, MȦ 6
f = 0 MȦ = 0 = M
f
Ȧ

R 6= 0 [Ir ], [Irc ], [IIr ], [Dr ], [IIr ], [Dr ] [Dr ]

R=0 [IIIr ], [−] [IIIr ], [−] [−]

Table 4: Petrov-Penrose types of SD Weyl spinor in para-Hermite spaces via properties

of the congruences of the SD null strings.

Types in CR
Cosmological constant MȦ 6= 0, MȦ 6
f = 0 MȦ = 0, MȦ 6
f = 0 MȦ = 0 = M
f
Ȧ

Λ 6= 0 [D], [−] not allowed [D]

Λ=0 [−] [−]
Types in UR
Cosmological constant MȦ 6= 0, MȦ 6
f = 0 MȦ = 0, MȦ 6
f = 0 MȦ = 0 = M
f
Ȧ

Λ 6= 0 [Dr ], [−] not allowed [Dr ]

Λ=0 [−] [−]

Table 5: Petrov-Penrose types of SD Weyl spinor in Einstein para-Hermite spaces via

properties of the congruences of the SD null strings.

5 Spaces equipped with three or four congruences of

SD null strings
5.1 General analysis
In this section we investigate the spaces which admit even richer structure then the
para-Hermite spaces, i.e. we consider the spaces admitting three or four distinct con-
gruences of SD null strings. As a starting point we take (complex or real) para-Hermite
space equipped with two distinct congruences Σ and Σ e (generated by the spinors mA and
A
µA , respectively, such that µ mA = 1). Consider now another cns, say Σ-congruence,
b
AB
Σ
b = sA sB S . Let Σ-congruence
b be complementary to the congruences Σ and Σ: e
A
Σ ∧ Σ 6= 0, Σ ∧ Σ 6= 0. Decomposing the spinor s according to the formula
b e b
m
sA = mmA + µµA = µ (ξmA + µA ) , ξ := (5.1)
µ

16
 Criteria Types in UR Types in CR
[Dr ]⊗[Dc ]
b>0 [2Z − 2Z̄]4(11) [2N1 − 2N ]2
[Dc ]⊗[Dc ]
b<0 [2R1s − 2R2t ]4(11) [2N1 − 2N ]2
[Dr ]⊗[Dr ]
[2R1nst − 2R2nst ]4(11)
[−]⊗[Nr ]
b = 0 BȦḂ 6= 0 [4Rn ]2(2) (2)
[4N ]a2
[−]⊗[−]
BȦḂ = 0 [4Rnst ]4(1) [4N ]1

Table 6: Possible types of the traceless Ricci tensor in para-Hermite spaces equipped with
two nonexpanding congruence of the SD null strings.

we find that m 6= 0 and µ 6= 0, otherwise the complementary conditions are not satisfied.
As a consequence of existence of the Σ-congruence
b we find
 
sA sB ∇AḂ sB = sA ∇AḂ ln ξ + 2ZAḂ − mA Mf + µA M = 0
Ḃ Ḃ (5.2)

Acting on (5.2) with sM ∇M Ḃ we find the integrability conditions of the equations (5.2)

2C (2) ξ 2 − 3C (3) ξ + 2C (4) = 0 (5.3)

According to Theorem 3.3 the spinor which generates the cns must be proportional to
a Penrose spinor. For the types [D] and [III] (in CR) or types [Dr ] and [IIIr ] (in UR)
we find, that Eq. (5.3) does not have any nonzero solutions (compare Table 3). For all
configurations of the C (i) coefficients which give the space of the type [II] (in CR) or [IIr ]
(in UR) we find that the equation (5.3) has exactly one nonzero solution. Indeed

2C (4) µ A
if C (2) = 0 then ξ = (3)
and sA = p (5.4)
3C 3C (3)
3C (3) µ A
if C (4) = 0 then ξ = and s A
= p
2C (2) 2C (2)
3C (3)
if δ = 0 then ξ = and sA = µpA
4C (2)
where pA is the third Penrose spinor (compare types [II] and [IIr ] in the Table 3).
However, if δ 6= 0 (in CR) or δ > 0 (in UR) the equation (5.3) has two different
nonzero solutions. One finds, that
1  (3) √ 
ξ± = 3C ± δ (5.5)
4C (2)
− ±
what implies sA = µp+A or sA = µpA where spinors pA are the third and fourth Penrose
spinor (compare types [I] and [Ir ] in Table 3).
Summing up: the ratio ξ can be found as a solution of the quadratic equation (5.3)
and it depends on the curvature coefficients of the SD Weyl spinor C (2) , C (3) and C (4) .
The space of the types [II] ⊗ [any] (in CR) or [IIr ] ⊗ [any] (in UR) admits three distinct
congruences of SD null strings, if
 
(ξmA + µA ) ∇AḂ ln ξ + 2ZAḂ − mA M f + µA M = 0
Ḃ Ḃ (5.6)

17
where ξ is given by (5.4). However, for the spaces of the types [I] ⊗ [any] (in CR) and
[Ir ] ⊗ [any] (in UR) there are two solutions ξ± . In this case we arrive at the equations
 
A A
(ξ+ m + µ ) ∇AḂ ln ξ+ + 2ZAḂ − mA MḂ + µA MḂ = 0
f (5.7a)
 
(ξ− mA + µA ) ∇AḂ ln ξ− + 2ZAḂ − mA M f + µA M = 0
Ḃ Ḃ (5.7b)

If (5.7a) are satisfied and (5.7b) are not (or vice versa), then the space of the types
[I] ⊗ [any] (in CR) or [Ir ] ⊗ [any] (in UR) admits three distinct congruences of SD null
strings. But if both (5.7a) and (5.7b) are satisfied, there are four distinct congruences of
SD null strings.
[Remark. Equation (5.3) has a solution for arbitrary ξ only if C (i) = 0, i = 2, 3, 4. It
means that arbitrary spinor sA generates the cns only if the spaces are left conformally
flat. Such spaces admit infinitely many cns.]
We gather the results of this section in two Theorems.

Theorem 5.1. If a complex (real) space with CABCD 6= 0 admits four distinct congruences
of SD null strings, then

(i) the space is of the type [I] ⊗ [any] ([Ir ] ⊗ [any])

(ii) all congruences are necessarily expanding

Proof. Spinors which generate distinct cns are mutually linearly independent. The ex-
istence of four distinct cns implies the existence of four mutually linearly independent
spinors and all of them are Penrose spinors (compare Theorem 3.3). Consequently the
space must be of the type [I] ⊗ [any] ([Ir ] ⊗ [any]). It proves (i). According to Theorem
3.4 the existence of nonexpanding cns implies that the space is algebraically degenerated
what contradicts (i). It proves (ii). 

Theorem 5.2. If a complex (real) space with CABCD 6= 0 admits three distinct congru-
ences of SD null strings and one of them is nonexpanding, then

(i) the space is of the type [II] ⊗ [any] ([IIr ] ⊗ [any])

(ii) curvature scalar R is necessarily nonzero R 6= 0

Proof. Nonexpanding cns is generated by the multiple Penrose spinor (compare Theorem
3.4). Moreover, all spinors which generate distinct cns are mutually linearly independent
and they are Penrose spinors (compare Theorem 3.3). The existence of three distinct cns
(one of them being nonexpanding) implies the existence of three Penrose spinors (one of
them is double). Consequently the space must be of the type [II] ⊗ [any] ([IIr ] ⊗ [any]).
It proves (i). (ii) follows immediately from Theorem 3.4. 
Possible Petrov-Penrose types of the spaces equipped with three complementary cns
are listed in the Table 7.

5.2 The metric of the spaces equipped with three distinct con-
gruences of SD null strings
The general metric of the space equipped with three distinct cns has been found
in [27]. The authors considered a para-Hermite space equipped with the congruences Σ

18
 Types in CR
Curvature scalar MȦ 6= 0, MȦ 6
f = 0, MȦ 6
c = 0 MȦ = 0, MȦ 6
f = 0, MȦ 6
c = 0
R 6= 0 [I], [II], [−] [II]
R=0 [−]
Types in UR
Curvature scalar MȦ 6= 0, MȦ 6
f = 0, MȦ 6
c = 0 MȦ = 0, MȦ 6
f = 0, MȦ 6
c = 0
R 6= 0 [Ir ], [IIr ], [−] [IIr ]
R=0 [−]

Table 7: Petrov-Penrose types of SD Weyl spinor in spaces equipped with three distinct
congruences of SD null strings via properties of these congruences.

e given by the Pfaff systems dz A = 0 and dz Ȧ = 0, respectively. It is well known [7]

and Σ,
that the metric of such a space can be brought to the form ds2 = 2fAḂ dz A dz Ḃ where
fAḂ = fAḂ (z M , z Ṁ ) are arbitrary complex holomorphic (real smooth) functions. The
third congruence, Σ b implies the existence of the functions f Ȧ such that this congruence
is defined by the Pfaff system df Ȧ = 0. It means that the alternate form of the metric
reads ds2 = 2ωAḂ dz A df Ḃ , where ωAḂ = ωAḂ (z M , f Ṁ ). Of course
ds2 = 2fAḂ dz A dz Ḃ = 2ωAḂ dz A df Ḃ (5.8)
Treating f Ȧ as a functions of (z A , z Ḃ ) we arrive at the equations
∂f Ṁ
fAḂ = ωAṀ (5.9a)
∂z Ḃ
∂f Ḃ
ω =0 (5.9b)
∂z (M A)Ḃ
Eqs. (5.9b) can be easily solved
∂fḂ
ωAḂ = Ω (5.10)
∂z A
where Ω = Ω(z A , z Ḃ ) is arbitrary function. Putting (5.10) into (5.9a) we arrive at the
metric [21]
∂f ∂f Ṁ A Ḃ
ds2 = 2Ω Ṁ dz dz (5.11)
∂z A ∂z Ḃ
Generally, the metric (5.11) admits three distinct expanding cns and its SD Weyl spinor
is of the types [I] or [II] in CR ([Ir ] or [IIr ] in UR). This metric cannot be Einsteinian.
The metric depends on three functions Ω and f Ḃ of the four variables (z A , z Ḃ ). Note,
that all congruences are complementary to each other so Σ ∧ Σ e 6= 0, Σ ∧ Σ
b 6= 0 and
e ∧Σ
Σ b 6= 0. It follows that

∂f Ȧ ∂fȦ ∂f Ȧ ∂fȦ
6= 0 , 6= 0 (5.12)
∂z B ∂zB ∂z Ḃ ∂zḂ
The special case with one of these congruences being nonexpanding has not been
considered in [27]. In this case we arrive at the following theorem:

19
Theorem 5.3. If a complex (real) space admits three distinct congruences of SD null
strings and one of them is nonexpanding, then the metric of this space can be brought to
the form
∂f ∂f Ṁ A Ḃ
ds2 = 2Ω Ṁ dz dz (5.13)
∂z A ∂z Ḃ
where Ω = Ω(z Ȧ ), f Ṁ = f Ṁ (z A , z Ḃ ) and (z A , z Ḃ ) are local coordinates.

Proof. The space admits three distinct cns, so the metric has the form (5.11). Detailed
∂ω
calculations prove, that the expansion of the Σ-congruence
b is proportional to the ∂zMMṄ .
Consider now Σ-congruence
b as a nonexpanding one. We have
∂ωM Ṅ
=0 (5.14)
∂zM
Feeding (5.14) with (5.10) one obtains

∂Ω ∂fḂ
=0 (5.15)
∂zA ∂z A
∂f Ḃ
Contracting (5.15) with ∂z M
we find

∂Ω ∂f Ȧ ∂fȦ
=0 (5.16)
∂z M ∂z B ∂zB
Finally, because of the (5.12) one arrives at the formula

∂Ω
M
= 0 ⇐⇒ Ω = Ω(z Ṅ ) (5.17)
∂z


5.3 The metric of the spaces equipped with four distinct con-
gruences of SD null strings
Four distinct cns are admitted only by the spaces with algebraically general SD Weyl
spinor. We assume the existence of congruences Σ, Σ, e Σ b and Σb given by the Pfaff
b
A Ȧ A Ȧ
systems dz = 0, dz = 0, df = 0 and df = 0, respectively. All those congruences are
complementary what implies

Σ∧Σ
e 6= 0 , Σ∧Σ
b 6= 0 , Σ∧Σ
b 6= 0 , e ∧Σ
Σ b 6= 0 , e ∧Σ
Σ b 6= 0 , b ∧Σ
Σ b 6= 0 (5.18)
b b b

Writing f A = f A (z B , z Ċ ) and f Ȧ = f Ȧ (z B , z Ċ ), from (5.18) we find that

∂f M ∂fM ∂f M ∂fM ∂f Ṁ ∂fṀ

α1 := 6= 0 , α2 := 6= 0 , α3 := 6= 0 (5.19)
∂z N ∂zN ∂z Ṅ ∂zṄ ∂z N ∂zN
∂f Ṁ ∂fṀ ∂f A ∂fA ∂f Ḃ ∂fḂ
α4 := 6= 0 , α1 α4 + α2 α3 − 4 M 6= 0 (5.20)
∂z Ṅ ∂zṄ ∂z ∂z Ṅ ∂zM ∂zṄ

Then we arrive at the following

20
Theorem 5.4. If a complex (real) space admits four distinct congruences of SD null
strings, then the metric of this space can be brought to the form
!
M Ṁ
∂f ∂f M ∂f ∂f
ds2 = 2Ω α3 − α1 Ṁ
dz A dz Ḃ (5.21)
∂z A ∂z Ḃ ∂z A ∂z Ḃ

where (z A , z Ḃ ) are local coordinates, Ω = Ω(z A , z Ḃ ) and the functions f M = f M (z A , z Ḃ )

and f Ṁ = f Ṁ (z A , z Ḃ ) satisfy the set of three equations

∂fA ∂fḂ ∂f A ∂f Ḃ
=0 (5.22)
∂z M ∂zM ∂z (Ṙ ∂z Ṡ)
Proof. Consider the congruences Σ and Σ. e Analogously to (5.8) the metric reads ds2 =
2fAḂ dz A dz Ḃ . Alternatively, choosing Σ
b and Σ-congruences as a basic ones, we have
b
b
2 A Ḃ
ds = 2ωAḂ df df . Of course

2fAḂ dz A dz Ḃ = 2ωAḂ df A df Ḃ (5.23)

what leads to the equations
!
∂f M ∂f Ṅ ∂f M ∂f Ṅ
fAḂ = ωM Ṅ + (5.24a)
∂z A ∂z Ḃ ∂z Ḃ ∂z A
∂f M ∂f Ṅ
ωM Ṅ =0 (5.24b)
∂z (A ∂z B)
∂f M ∂f Ṅ
ωM Ṅ =0 (5.24c)
∂z (Ȧ ∂z Ḃ)
From (5.24b) we obtain
∂f M ∂fṄ
ωM Ṅ = ω (5.25)
∂z A ∂z A
∂fN
where ω = ω(z A , z Ḃ ) is some function. Contracting (5.25) with ∂zA
one gets
1 ∂f ∂fM
α1 ωM Ṅ = ω Ṅ (5.26)
2 ∂z A ∂zA
Remembering, that α1 6= 0 and re-defining function ω we find the solution for ωM Ṅ
∂fM ∂fṄ
ωM Ṅ = −2Ω (5.27)
∂z A ∂zA
Putting (5.27) into (5.24a) one finds the solution for fAḂ and finally arrives at the metric
(5.21). The last step is to feed (5.24c) by (5.27). The result is the system (5.22). 
[Remark. One can first solve the Eq. (5.24c). In this case we arrive at the equivalent
form of the metric
!
M Ṁ
∂f ∂f M ∂f ∂f
ds2 = 2Ω1 α4 − α2 Ṁ
dz A dz Ḃ (5.28)
∂z A ∂z Ḃ ∂z A ∂z Ḃ

with the equations

∂fM ∂fṄ ∂f M ∂f Ṅ
=0 ] (5.29)
∂z Ȧ ∂zȦ ∂z (A ∂z B)

21
6 Concluding remarks
There are several directions of further investigations. In this paper we considered
only SD congruences. The manifold M can be equipped with similar structures on ASD
side. Intersections of these congruences constitute the sngc. Because in CR and UR,
SD and ASD structures are unrelated one can consider the spaces with mixed Petrov-
Penrose types (like [D] ⊗ [N] or [IIrc ] ⊗ [IIIr ]) equipped with many sngc with different
properties of optical scalars (twist, expansion). Note, that the types [I] ⊗ [I] (in CR) and
[Ir ] ⊗ [Ir ] (in UR) admits 4 SD and 4 ASD congruences of null strings. Intersections of
these congruences give 16 distinct sngc. Do the spaces having so rich structure exist? Or
maybe some (unrecognized yet) conditions make such spaces impossible to exist?
There are 6 different types of SD Weyl spinor in CR and 10 in UR (see [5, 26]).
The existence of cns allows to distinguish the special subtypes. For example, type [I] in
general do not admit any cns. More special type [I]e admits one cns (superscript e means
that the corresponding cns is expanding while n denotes nonexpanding cns). Types [I]ee ,
[I]eee and [I]eeee are equipped with two, three and four cns, correspondingly. All possible
subtypes of CABCD in CR and UR are listed on the Schemes 1 and 2. ([−]e means, that
all cns are expanding, while [−]n means, that at least one of them in nonexpanding).

0 cns: [I] [II] [III] [D] [N] [−]

        
1 cns: [I]e [II]e [II]n [III]e [III]n [D]e [D]n [N]e [N]n

          
2 cns: [I]ee [II]ee [II]en [III]ee [III]en [D]ee [D]en [D]nn

   
3 cns: [I]eee [II]eee [II]een


4 cns: [I]eeee

 
∞ cns: [−]e [−]n

Scheme 1: Types of CABCD in CR equipped with different numbers of congruences of

the SD null strings.

The possible applications of our results concern not only CR and UR but also HR.
In [13] there is an example of Lorentzian manifold equipped with three distinct sngc
(which are the intersection of 3 SD and 3 ASD congruences of null strings). The general
form of the metric of the space equipped with three distinct sngc is still unknown. The
richest possible structure appears as intersections of 4 SD and 4 ASD congruences of null
strings. In HR it gives 4 distinct sngc. In [31] Trautman asks: ”In dimension 4, are
there not conformally flat Lorentz manifolds that have 4 distinct shear-free null geodesic
congruences?”. Such spaces (if exist) cannot be conformal to Einstein spaces (because of
the Goldberg - Sachs theorem). Our considerations (Theorem 3.3) give more properties
of such spaces: they must be of Petrov-Penrose type [I]. The explicit metric as well as the

22
 0 cns: [Ir ] [Irc ] [Ic ] [IIr ] [IIrc ] [IIIr ] [Dr ] [Dc ] [Nr ] [−]

           
1 cns: [Ir ]e [Irc ]e [IIr ]e [IIr ]n [IIrc ]e [IIrc ]n [IIIr ]e [IIIr ]n [Dr ]e [Dr ]n [Nr ]e [Nr ]n

           
2 cns: [Ir ]ee [Irc ]ee [IIr ]ee [IIr ]en [IIIr ]ee [IIIr ]en [Dr ]ee [Dr ]en [Dr ]nn

   
3 cns: [Ir ]eee [IIr ]eee [IIr ]een


4 cns: [Ir ]eeee

 
∞ cns: [−]e [−]n

Scheme 2: Types of CABCD in UR equipped with different numbers of congruences of

the SD null strings.

possible types of tensor of matter admitted by such spaces are unknown. The problem of
Lorentzian manifolds with 4 distinct sngc is very interesting and we are going to analyze
it in details in near future.
Quite surprisingly, the space with two distinct expanding cns seems to admit all
algebraic types of the traceless Ricci tensor. We believe, that in fact there are some
restrictions which reduce the number of possible types in such a case. Similarly, in spaces
equipped with three or four expanding cns those restrictions should be even stronger.
Frankly, this problem is more advanced and it involves deeper analysis.
Finally, expansion of the cns is the most transparent property of such structure and
has clear geometrical interpretation. It seems, that the Sommers vector carries some in-
formation about cns. Natural question arises if the properties the Sommers vector define
a new subclassification of cns?

Acknowledgments
Some points of the present paper were presented in July 2016 in Brno at the 13th con-
ference Differential Geometry and its Applications and in September 2016 in Krakow at
The 3rd Conference of the Polish Society on Relativity. The author is indebted to Maciej
Przanowski for his interest in this work and for help in many crucial matters.

References
[1] Alekseevsky D.V, Medori C, and Tomassini A., Homogeneous para-Kähler Einstein
manifolds, Russian Mathematical Surveys, Volume 64, Number 1 (2009)

[2] Boyer C.P., Finley III J.D. and Plebański J.F, Complex generel relativity, H and HH
spaces - a survey of one approach, in General Relativity and Gravitation. Einstein
Memorial Volume, ed. A. Held (Plenum, New York, 1980) vol. 2. pp. 241-281 (1980)

[3] Chaichi M., Garcı́a - Rı́o E. and Matsushita Y., Curvature properties of four -
dimensional Walker metrics, Class. Quantum Grav., 22, 559 (2005)

23
 [4] Chudecki A., On some examples of para-Hermite and para-Kähler Einstein spaces
with Λ 6= 0, Journal of Geometry and Physics 112, 175–196
[5] Chudecki A., Classification of the traceless Ricci tensor in 4-dimensional pseudo-
Riemannian spaces of neutral signature, Acta Physica Polonica B, Vol. 48, No. 1,
53-74
[6] Chudecki A. and Przanowski M., From hyperheavenly spaces to Walker and Osser-
man spaces: I, Class. Quantum Grav., 25 145010 (2008)
[7] Flaherty E. J., Hermitian and Kählerian Geometry in Relativity, Lecture Notes in
Physics, Springer-Verlag, Berlin (1976)
[8] Goldberg J.N. and Sachs R.K., A theorem on Petrov types, Acta Physica Polonica
(supplement), 22, 13 (1962)
[9] Law P.R. and Matsushita Y., Real AlphaBeta-geometries and Walker geometry,
Journal of Geometry and Physics, 65 35-44 (2013)
[10] Nurowski P. and An D., Twistor Space for Rolling Bodies, Communications in Math-
ematical Physics, 326, 2, 393-414 (2013)
[11] Nurowski P., Bor G. and Lamoneda L.H., The dancing metric, G2-symmetry and
projective rolling, arXiv:1506.00104
[12] Nurowski P. and Taghavi-Chabert A., A Goldberg-Sachs theorem in dimension three,
Class. Quantum Grav. 32, 115009 (2015)
[13] Nurowski P. and Trautman A., Robinson manifolds as the Lorentzian analogs of
Hermite manifolds, Differential Geometry and its Applications, 17, 175-195 (2002)
[14] Penrose R., A spinor approach to general relativity, Ann. Phys. 10, 171 (1960).
[15] Pirani F.A.E., Introduction to Gravitational Radiation Theory, in Lectures in Gen-
eral Relativity Theory, Brandeis Summer Institute, Prentice - Hall, Inc. (1964)
[16] Plebański J. F., Spinors, tetrads and forms, unpublished monograph of Centro de
Investigacion y Estudios Avanzados del IPN, Mexico 14 (1974)
[17] Plebański J.F., The algebraic Structure of the tensor of matter, Acta Physica
Polonica, vol. 26, 963-1020 (1964)
[18] Plebański J. F., The spinorial and helicity formalisms of Riemannian structures in
complex and real four dimensions, preprint of Centro de Investigación y de Estudios
Avanzados del IPN, México D.F., México (1980)
[19] Plebański J. F. and Hacyan S., Null geodesic surfaces and Goldberg - Sachs theorem
in complex Riemannian spaces, J. Math. Phys., 16, 2403 (1975 )
[20] Plebański J. F. and Przanowski M., Null tetrad, spinor and helicity formalisms for all
4-dimensional Riemannian manifolds. I. Null tetrads and spinors, Acta Phys. Polon.,
B19 805 (1988)
[21] Plebański J. F. and Rózga K. The optics of null strings, J. Math. Phys., 25, 1930
(1984)

24
[22] Przanowski M., Remarks concerning the geometry of null strings, Acta Phys. Polon.,
B11 945 (1979)

[23] Przanowski M., Formański S. and Chudecki A., Notes on para-Hermite-Einstein

spacetimes, International Journal of Geometric Methods in Modern Physics, Vol.9,
No.1, 1250008 (2012)

[24] Przanowski M. and Plebański J.F., Generalized Goldberg-Sachs theorems in complex

and real space-times. I., Acta Physica Polonica, Vol. B10 (1979)

[25] Przanowski M. and Plebański J.F., Generalized Goldberg-Sachs theorems in complex

and real space-times II, Acta Phys. Polon. B 10, 573 (1979)

[26] Rod Gover A., Denson Hill C. and Nurowski P., Sharp version of the Goldberg-Sachs
theorem, Annali di Matematica, Vol. 190, Issue 2, pp 295–340 (2011)

[27] Robinson I. and Rózga K. Congruences of null strings in complex space-times and
some Cauchy-Kovalevski-like problems, J. Math. Phys., 25, 1941 (1984)

[28] Stephani H., Kramer D., MacCallum M. A. H., Hoenselaers C. and Herlt E., Exact
Solutions to Einstein’s Field Equations. Second Edition, Cambridge University Press,
Cambridge (2003)

[29] Szereszewski A. and Tafel J., Solutions of the Rarita–Schwinger equation in Einstein
spaces, Physics Letters A 297, 359–362 (2002)

[30] Taghavi-Chabert A., The complex Goldberg-Sachs theorem in higher dimensions, J.

Geom. Phys. 62, no. 5, 981-1012 (2012)

[31] Trautman A., Robinson manifolds and Cauchy-Riemann spaces, Classical and Quan-
tum Gravity, 19, R1 (2002)

[32] Walker A.G., Canonical form for a Riemannian space with a parallel field of null
planes, Quart. J. Math. Oxford (2), 1, 69 (1950)

25