Continuous order-to-order quantum phase transitions from fixed-point annihilation

David J. Moser and Lukas Janssen

Institut für Theoretische Physik and Würzburg-Dresden Cluster of Excellence ct.qmat, TU Dresden, 01062 Dresden, Germany
(Dated: December 11, 2024)
A central concept in the theory of phase transitions beyond the Landau-Ginzburg-Wilson paradigm is frac-
tionalization: the formation of new quasiparticles that interact via emergent gauge fields. This concept has
been extensively explored in the context of continuous quantum phase transitions between distinct orders that
break different symmetries. We propose a mechanism for continuous order-to-order quantum phase transitions
that operates independently of fractionalization. This mechanism is based on the collision and annihilation
of two renormalization group fixed points: a quantum critical fixed point and an infrared stable fixed point.
The annihilation of these fixed points rearranges the flow topology, eliminating the disordered phase associated
with the infrared stable fixed point and promoting a second critical fixed point, unaffected by the collision, to a
arXiv:2412.06890v1 [cond-mat.str-el] 9 Dec 2024

quantum critical point between distinct orders. We argue that this mechanism is relevant to a broad spectrum of
physical systems. In particular, it can manifest in Luttinger fermion systems in three spatial dimensions, leading
to a continuous quantum phase transition between an antiferromagnetic Weyl semimetal state, which breaks
time-reversal symmetry, and a nematic topological insulator, characterized by broken lattice rotational sym-
metry. This continuous antiferromagnetic-Weyl-to-nematic-insulator transition might be observed in rare-earth
pyrochlore iridates 𝑅2 Ir2 O7 . Other possible realizations include kagome quantum magnets, quantum impurity
models, and quantum chromodynamics with supplemental four-fermion interactions.

The exploration of quantum phase transitions beyond the nal control parameter 𝑁. For 𝑁 larger than a critical value 𝑁c ,
Landau-Ginzburg-Wilson paradigm [1, 2] has garnered signif- there is an interacting disordered phase, which is separated
icant attention in recent years. A particular focus has been from two ordered phases by continuous phase transition lines,
on continuous quantum phase transitions between phases that see Fig. 1(a). As 𝑁 decreases, the fixed point associated with
break different symmetries [3]. Traditionally, it has been be- the disordered phase (D) and those governing one of the two
lieved that such transitions require fractionalization, where transition lines (QCP2 ) approach each other and eventually
new “deconfined” quasiparticles with fractional quantum num- collide at a critical value 𝑁c . For 𝑁 smaller than 𝑁c , Fig. 1(b),
bers arise and interact via emergent gauge fields. The in- the fixed-point annihilation alters the flow topology by remov-
tense study of deconfined quantum phase transitions in recent ing the disordered phase, leaving behind a continuous phase
years has significantly advanced our understanding of quan- transition line between two phases with distinct orders.
tum critical phenomena. Notable insights include the discov-
ery of emergent symmetries [4–7], the identification of spec-
troscopic signatures of fractionalization [8], the uncovering
of unexpected webs of field-theoretical dualities [9], and the (a) O
2
𝑁 > 𝑁c (b) O
2
𝑁 < 𝑁c
observation of anomalous scaling behaviors in entanglement
entropy [10, 11]. However, the most recent numerical studies QCP2
suggest that the deconfined quantum phase transition, at least
𝐺2

in its original formulation within SU(2) spin models [1–3], is D

ultimately first-order, characterized by a finite, though large,
correlation length at the transition point [6, 11–14]. QCP1 QCP1
Here, we propose a mechanism that provides a different O1 O1
pathway to continuous order-to-order quantum phase transi- 𝐺1 𝐺1
tions, one that does not rely on fractionalization. Our proposal
involves the collision and annihilation of two renormalization FIG. 1. Mechanism for continuous order-to-order transition from
fixed-point annihilation. (a) Schematic RG flow in the space spanned
group (RG) fixed points: a quantum critical fixed point and an by two couplings, 𝐺 1 and 𝐺 2 , for 𝑁 > 𝑁c . Arrows indicate flow
infrared stable fixed point. We argue that the mechanism is towards infrared. Beyond an interacting disordered phase (gray)
relevant to a broad spectrum of physical systems. In particular, at small couplings, governed by the fixed point D, there are two
we explicitly demonstrate within a RG calculation that three- ordered phases (red and blue) in the strong-coupling regime, governed
dimensional Luttinger fermion systems undergo a continuous by the fixed points O1 and O2 . The disordered phase is separated
quantum phase transition between an antiferromagnetic Weyl from the two ordered phases by continuous phase transition lines
semimetal state, marked by broken time-reversal symmetry, (purple), governed by the quantum critical fixed points QCP1 and
QCP2 . As 𝑁 decreases, the fixed points D and QCP2 approach
and a nematic topological insulator state, characterized by bro- each other and eventually collide at a critical value 𝑁c . (b) Same
ken lattice rotational symmetry. Further possible realizations as (a), but for 𝑁 < 𝑁c . The fixed-point annihilation at 𝑁c alters the
in other setups are discussed as well. flow topology, eliminating the disordered phase and leaving behind a
The mechanism is schematically illustrated in Fig. 1, which single continuous phase transition line (purple) between two phases
shows RG flow diagrams for two different values of an exter- with distinct orders.
 2

FIG. 2. (a) Phase diagram of Luttinger model in the low-temperature limit as function of short-range couplings 𝐺 1 and 𝐺 2 for fixed representative
values of the charge 𝑒 2 = 3𝜋 2 Λ/2 and the anisotropy parameter 𝛿 = −1/2 from mean-field theory, which becomes exact in the limit 𝑁 → ∞.
The Luttinger-Abrikosov-Beneslavskii (LAB) phase at small couplings realizes a three-dimensional non-Fermi liquid (gray). For sufficiently
large 𝐺 1 , a Weyl semimetal emerges, characterized by all-in-all-out (AIAO) antiferromagnetic order on the pyrochlore lattice. For sizable 𝐺 2 , a
nonmagnetic nematic topological insulator is stabilized. Color scale indicates magnitudes of AIAO order parameter ⟨𝜙⟩ ∝ ⟨𝜓 † 𝛾45 𝜓⟩ (red) and
nematic order parameter ⟨𝜑⟩ ∝ ⟨𝜓 † 𝛾5 𝜓⟩ (blue). Single (double) lines indicate continuous (discontinuous) phase transitions. (b) Same as (a),
but for 𝑁 = 10 from RG analysis, qualitatively agreeing with the mean-field result. (c) Same as (b), but for 𝑁 = 2. The LAB phase shrinks in
favor of the nematic phase as the LAB fixed point and the quantum critical fixed point associated with the nematic instability approach each
other. (d) Same as (b), but for 𝑁 = 1, corresponding to the case relevant for the pyrochlore iridates 𝑅2 Ir2 O7 . The LAB phase is removed as
a consequence of the fixed-point annihilation, leaving behind a continuous phase transition line between the AIAO antiferromagnetic Weyl
semimetal and the nematic topological insulator for small 𝐺 2 .

Pyrochlore iridates. We begin by reviewing the interacting the summation symbols over 𝑖 = 1, . . . , 𝑁. The long-range
Luttinger fermion model in three spatial dimensions [15–18]. tail parameterized by the charge 𝑒 2 must be taken into account
This continuum model effectively describes electronic excita- because the Coulomb interaction is only marginally screened
tions near a quadratic band touching point at the Fermi level, due to the vanishing density of states at the Fermi level [24].
as they occur in pyrochlore iridates 𝑅2 Ir2 O7 , with 𝑅 a rare- The coupling 𝐺 1 originates from local Hubbard repulsion [28],
earth element [19]. In the low-temperature limit, the model and is believed to increase with decreasing radius of the rare-
can be defined by the Euclidean action 𝑆 = 𝑆0 + 𝑆int , where 𝑆0 earth ion [19]. The coupling 𝐺 2 may be assumed to be small
corresponds to the noninteracting part, on the microscopic level. However, as will be shown below,
∫ ! 𝐺 2 is generated by 𝐺 1 and the long-range interaction, and we
∑︁
𝑁 ∑︁5
𝑆0 = d𝜏 d3 𝑥® †
𝜓 𝑖 𝜕𝜏 + (1 + 𝑠 𝑎 𝛿)𝑑 𝑎 (−i∇)𝛾 𝑎 𝜓𝑖 , therefore include it from the outset.
𝑖=1 𝑎=1
(1) In the large-𝑁 limit and for small 𝐺 1 and 𝐺 2 , the cou-
plings flow to an infrared stable RG fixed point, characterizing
which originates in the Luttinger Hamiltonian [20]. In the an interacting semimetal phase. This phase, which realizes a
above, 𝑁 corresponds to the number of band touching points three-dimensional non-Fermi liquid, is known as the Luttinger-
at the Fermi level, with 𝑁 = 1 in the case of the pyrochlore Abrikosov-Beneslavskii (LAB) phase [15–17]. Sizable four-
iridates, 𝜓𝑖 is a four-component Grassmann field, 𝑠 𝑎 B 1 fermion interactions beyond a finite threshold, however, drive
(𝑠 𝑎 B −1) for 𝑎 = 1, 2, 3 (𝑎 = 4, 5), 𝛿 parametrizes the instabilities of the LAB state. Figure 2(a) shows the mean-field
cubic anisotropy with −1 ≤ 𝛿 ≤ 1, the 4 × 4 matrices 𝛾 𝑎 phase diagram as function of 𝐺 1 and 𝐺 2 for fixed representative
fulfill the Euclidean
√ Clifford algebra,
√ {𝛾 𝑎 , 𝛾𝑏 } = 2𝛿√ 𝑎𝑏 1, and values of 𝛿 and 𝑒 2 . Technical details of the mean-field calcu-
® B 3𝑝 𝑦 𝑝 𝑧 , 𝑑2 ( 𝑝)
𝑑1 ( 𝑝) ® B 3𝑝 𝑥 𝑝 𝑧 , 𝑑3 ( 𝑝) ® B 3𝑝 𝑥 𝑝 𝑦 , lation, which becomes exact in the limit 𝑁 → ∞, are provided
√
® B 3( 𝑝 2𝑥 − 𝑝 2𝑦 )/2, and 𝑑5 ( 𝑝)
𝑑4 ( 𝑝) ® B (2𝑝 2𝑧 − 𝑝 2𝑥 − 𝑝 2𝑦 )/2 in the Supplemental Material (SM) [29]. Beyond the LAB
are proportional to the ℓ = 2 real spherical harmonics. phase at small couplings, we identify two ordered phases: For
We account for both short-range repulsion and the long- sufficiently large 𝐺 1 , a Weyl semimetal phase emerges, char-
range tail of the Coulomb interaction via [21–27] acterized by the Ising order parameter ⟨𝜙⟩ ∝ ⟨𝜓 † 𝛾45 𝜓⟩, which
∫  signals spontaneous time-reversal symmetry breaking. On the
𝐺 2 ∑︁ †
5
3 𝐺1 † pyrochlore lattice, this phase exhibits all-in-all-out (AIAO)
𝑆int = − d𝜏 d 𝑥® (𝜓 𝛾45 𝜓) 2 + (𝜓 𝛾 𝑎 𝜓) 2
2𝑁 2𝑁 𝑎=1 antiferromagnetic order [23, 26, 27]. In contrast, for sizable
∫  𝐺 2 , we find a nonmagnetic phase that breaks lattice rotational
𝑒2 3 1
+ d 𝑦® 𝜓 ( 𝑥®)𝜓( 𝑥®)
†
𝜓 ( 𝑦®)𝜓( 𝑦®) , (2)
†
symmetry. This phase is characterized by the nematic order
8𝜋𝑁 |®
𝑥 − 𝑦® |
parameter ⟨𝜑⟩ ∝ ⟨𝜓 † 𝛾5 𝜓⟩, constructed from one of the five
where we defined 𝛾45 B i𝛾4 𝛾5 . For notational simplicity, we irreducible components of traceless symmetric tensors in three
abbreviated 𝜓 ≡ 𝜓𝑖 (𝜏, 𝑥®) and 𝜓(®
𝑥 ) ≡ 𝜓𝑖 (𝜏, 𝑥®), and suppressed dimensions [22]. It corresponds to a nematic topological in-
 3

sulator [21, 25]. While the transitions from the LAB phase
to the ordered phases are continuous, the direct transition be-
tween the two ordered phases is first order, consistent with the
Landau-Ginzburg-Wilson paradigm.
RG analysis. At finite 𝑁, fluctuation effects can become
significant. To account for these effects, we use an RG anal-
ysis in the dynamical bosonization framework [30–32]. This
approach applies a Hubbard-Stratonovich transformation after
each RG step, effectively capturing potential nonperturbative
effects from the RG-induced generation of four-fermions cou-
plings while avoiding the unphysical singularities associated
with their runaway flow. The method has been successfully FIG. 3. (a) Normalized order parameters (OPs) ⟨𝜑⟩/𝐴 𝜑 and ⟨𝜙⟩/𝐴 𝜙
applied to Luttinger semimetals [25], demonstrating the exis- for nematic (blue) and AIAO (red) order, respectively, for 𝑁 = 1
tence of a critical value 𝑁c of quadratic band touching points at as functions of the ultraviolet value of the short-range coupling 𝐺 1
the Fermi level, consistent with earlier analyses [21, 24]. For from RG analysis for representative fixed ultraviolet values of the
short-range coupling 𝐺 2 = 0, charge 𝑒 2 = 3𝜋 2 Λ/2, and anisotropy
𝑁 ↘ 𝑁c , the infrared stable fixed point corresponding to the
parameter 𝛿 = −1/2. There is a direct continuous transition at 𝐺 1c =
LAB phase collides and annihilates with the quantum critical 0.565·2𝜋 2 /Λ between the nematic topological insulator for 𝐺 1 < 𝐺 1c
fixed point that governs the transition between the LAB phase and the AIAO Weyl semimetal for 𝐺 1 > 𝐺 1c . The normalization
and the nematic topological insulator. All the different ap- factors were chosen such that ⟨𝜙⟩ = 𝐴 𝜙 and ⟨𝜑⟩ = 𝐴 𝜑 for 𝐺 1 −𝐺 1c =
proaches predict a value of 𝑁c around 2, and therewith above ±0.1 · 2𝜋 2 /Λ, respectively. The order parameters scale as ⟨𝜙⟩ ∝
the physical case 𝑁 = 1 relevant for the pyrochlore iridates. (𝐺 1 −𝐺 1c ) 𝛽 𝜙 and ⟨𝜑⟩ ∝ (𝐺 1c −𝐺 1 ) 𝛽 𝜑 for 𝐺 1 approaching 𝐺 1c from
Crucially, the dynamically bosonized RG approach allows above and below, respectively, with the same exponent 𝛽 𝜙 = 𝛽 𝜑 = 2,
us to compute the order parameters as functions of the micro- but with significantly different prefactors. (b) Finite-temperature
phase diagram for 𝑁 = 1 as function of 𝐺 1 from RG analysis, with
scopic couplings by integrating out the RG flow. This enables ultraviolet values of 𝐺 2 , 𝑒 2 , and 𝛿 as in (a). The quantum critical
us to determine the nature of the phase transitions explicitly. point between the nematic topological insulator (blue) and the AIAO
Details of the method are provided in the SM [29]. For large Weyl semimetal (red) is marked as black dot. The slow RG flow
𝑁, we recover the infrared stable fixed point that governs the resulting from fixed-point annihilation causes a striking anisotropy
LAB phase. Additionally, we locate two other interacting fixed between the two sides of the transitions, with the critical temperature
points, each with one RG relevant direction and otherwise only significantly suppressed on the nematic side. This suppression of the
RG irrelevant directions. By integrating out the RG flow, we nematic order creates a broad crossover region dominated by LAB-
like behavior at finite temperature (gray).
identify these fixed points as the quantum critical points for
the transitions from the LAB phase towards the AIAO an-
tiferromagnetic Weyl semimetal and the nematic topological
insulator, respectively. The resulting quantum phase diagram leads to a continuous transition without fine tuning, provided
for a representative value of 𝑁 = 10 is shown in Fig. 2(b) and 𝐺 2 is not too large.
qualitatively aligns with the mean-field result in Fig. 2(a). As To illustrate this fact, we show in Fig. 3(a) the nematic order
𝑁 decreases, the LAB fixed point and the quantum critical parameter ⟨𝜑⟩ ∝ ⟨𝜓 † 𝛾5 𝜓⟩ and the AIAO antiferromagnetic
fixed point associated with the nematic instability approach order parameter ⟨𝜙⟩ ∝ ⟨𝜓 † 𝛾45 𝜓⟩ for 𝑁 = 1 as functions of the
each other. This implies that the LAB phase shrinks in the ultraviolet value of 𝐺 1 while keeping the ultraviolet values of
phase diagram in favor of the nematic phase, as exemplified 𝐺 2 , 𝑒 2 , and 𝛿 fixed. Details on obtaining the order parame-
for 𝑁 = 2 in Fig. 2(c). Upon further decreasing 𝑁, the two ters by integrating out the flow equations are provided in the
fixed points eventually collide at a critical value 𝑁c = 1.856, SM [29]. The two order parameters vanish continuously upon
in agreement with Ref. [25]. The fixed-point annihilation for approaching the quantum critical point at 𝐺 1c from above and
𝑁 < 𝑁c alters the flow topology by removing the LAB phase. below, respectively. The same behavior is observed for other
We emphasize that no additional fixed points are involved in values of 𝐺 2 , 𝑒 2 , and 𝛿, provided 𝐺 2 is not too large. There
the collision; in particular, the quantum critical fixed point as- is a direct continuous order-to-order quantum phase transition
sociated with the AIAO instability remains within the physical without an intermediate mixed phase and without additional
space of real couplings for all 𝑁 ≥ 1. The resulting quantum fine tuning. The corresponding quantum critical behavior is
phase diagram for the physical case of 𝑁 = 1, relevant for the governed by the RG flow near the fixed point associated with
pyrochlore iridates, is shown in Fig. 2(d). Importantly, there is the AIAO instability. In the present model, this fixed point
now a direct transition between the AIAO Weyl semimetal and is situated at extreme cubic anisotropy [23], allowing for the
the nematic topological insulator upon tuning a single param- exact computation of all critical exponents [27, 29]. We find
eter, 𝐺 1 . Although the two ordered states break different sym- that the anomalous dimensions for both the AIAO and nematic
metries, and the Landau-Ginzburg-Wilson paradigm would orders, respectively are 𝜂 𝜙 = 𝜂 𝜑 = 1, the correlation-length
predict a discontinuous transition, the presence of the quan- exponents 𝜈 𝜙 = 𝜈 𝜑 = 1, and the dynamical exponent 𝑧 = 2.
tum critical fixed point associated with the AIAO instability Assuming hyperscaling holds, we find that the two order pa-
 4

rameters scale as ⟨𝜙⟩ ∝ (𝐺 1 −𝐺 1c ) 𝛽 𝜙 and ⟨𝜑⟩ ∝ (𝐺 1c −𝐺 1 ) 𝛽 𝜑 (a) (b) (𝐽1 , 𝐽2 , 𝐽3 ) = (1, 𝛼, 𝛼)

as 𝐺 1 approaches 𝐺 1c from above and below, respectively, with 𝐽2
Quantum critical
the same exponent 𝛽 𝜙 = 𝛽 𝜑 = 2. In the SM [29], we replot the
𝐽3
𝐽1
data from Fig. 3 using a double-logarithmic scale, confirming

𝑇
this predicted scaling of the order parameters. DSL CSL

VBS T broken
Asymmetry in energy scales. Although the two order pa-
𝛼c ≃ 0.4
rameters scale with the same exponents, their correspond-
𝛼
ing prefactors differ significantly. Since the ratio of these
FIG. 4. (a) Kagome lattice, indicating first (𝐽1 ), second (𝐽2 ), and third
prefactors depends on microscopic parameters and is as such (𝐽3 ) neighbors. (b) Schematic finite-temperature phase diagram of
nonuniversal, it cannot be quantitatively predicted within our kagome Heisenberg model as function of 𝛼 = 𝐽2 /𝐽1 = 𝐽3 /𝐽1 , from
continuum model. However, at equal distances from critical- RG analysis of effective QED3 -Gross-Neveu model. The gapped
ity, we qualitatively find that the nematic order parameter for chiral spin liquid (CSL) breaks time reversal. The disordered gap-
𝐺 1 < 𝐺 1c is typically several order of magnitude smaller than less Dirac spin liquid (DSL) is unstable at low temperatures to-
the AIAO order parameter for 𝐺 1 > 𝐺 1c . This phenomenon wards a valence bond solid (VBS), leaving behind a continuous zero-
temperature transition between distinct orders.
arises as a direct consequence of the fixed-point annihilation
occuring at 𝑁c = 1.856, which is only slightly above the phys-
ical value of 𝑁 = 1. For 𝑁 just below 𝑁c , the complex by the Hamiltonian
LAB fixed point has a small imaginary part. As a result, ∑︁ ∑︁ ∑︁
the RG flow near the location in parameter space where the H = 𝐽1 𝑆®𝑖 · 𝑆® 𝑗 + 𝐽2 𝑆®𝑖 · 𝑆® 𝑗 + 𝐽3 𝑆®𝑖 · 𝑆® 𝑗 . (3)
fixed-point annihilation has occurred becomes logarithmically ⟨𝑖 𝑗 ⟩ ⟨⟨𝑖 𝑗 ⟩⟩ ⟨⟨⟨𝑖 𝑗 ⟩⟩⟩
slow. This slow flow leads to an exponential suppression of
the effective energy scale associated with the nematic order, a Here, 𝑆®𝑖 represents a spin-1/2 at site 𝑖, while ⟨𝑖 𝑗⟩, ⟨⟨𝑖 𝑗⟩⟩, and
phenomenon known as “walking behavior” in the field-theory ⟨⟨⟨𝑖 𝑗⟩⟩⟩ correspond to first-, second-, and third-neighbor bonds,
literature [33, 34]. Walking behavior is observed in fixed-point respectively, with 𝐽1 , 𝐽2 , and 𝐽3 denoting the associated ex-
annihilation scenarios across various systems, including quan- change couplings, see Fig. 4(a). The model describes mag-
tum impurity models [35–38], Potts models [39–42], Abelian netic Mott insulators crystallizing in a layered kagome struc-
Higgs models [43, 44], Wess-Zumino-Witten models [45–47], ture, such as herbertsmithite ZnCu3 (OH)6 Cl2 [57], kapella-
QED3 [48–51], and QCD4 [52–56]. site Cu3 Zn(OH)6 Cl2 [58], or YCu3 (OH)6.5 Br2.5 [59, 60]. For
dominant antiferromagnetic 𝐽1 > 0, numerical simulations at
The asymmetry in the effective energy scales on the two finite temperature or on finite lattices observe no evidence for
different sides of the continuous order-to-order transition is any long-range-ordered state, suggesting the formation of a
illustrated in the finite-temperature phase diagram shown in quantum spin liquid [61–69]. The nature of this state has been
Fig. 3(b). Details on obtaining the phase diagram are pro- a matter of significant debate, with recent simulations indicat-
vided in the SM [29]. On the nematic side of the transition, ing a gapless U(1) Dirac spin liquid [65–67]. For increasing
the critical temperature is significantly suppressed by several second- and third-neighbor interactions 𝐽2 and 𝐽3 , a transition
orders of magnitude in comparison with the AIAO side. This to a gapped chiral spin liquid, characterized by broken time re-
suppression of the nematic order creates a broad intermediate versal symmetry with finite scalar spin chirality ⟨𝑆®𝑖 · ( 𝑆® 𝑗 × 𝑆®𝑘 )⟩,
regime at finite temperature, separated from the nonuniver- has been found [70–74]. The transition has been argued to
sal high-energy regime by a crossover scale 𝑇★LAB . In this be continuous [70] and described by the QED3 -Gross-Neveu
regime, the RG flow is “stuck” in the vicinity of the, now model with 𝑁 = 2 four-component Dirac fields [65, 75–80].
imaginary, LAB fixed point [21, 25], and thermodynamic and As detailed in the SM [29], the QED3 -Gross-Neveu model
transport observables exhibit power laws with nontrivial ex- exhibits fixed-point annihilation as a function of 𝑁: At large
ponents, characteristic of a non-Fermi liquid [17]. While the 𝑁, an infrared stable fixed point associated with the conformal
asymmetry in the effective energy scales on either side of the phase of QED3 [48, 50, 81, 82] exists, describing the U(1)
transition decreases with increasing 𝐺 2 , we find it to remain Dirac spin liquid state. Additionally, there are two quantum
significant for all parameter sets that we have studied. This critical fixed points, associated with time-reversal-symmetry-
indicates that it can serve as a distinctive signature for experi- breaking and flavor-symmetry-breaking instabilities. The for-
mentally identifying the proposed mechanism for continuous mer governs the continuous transition between the U(1) Dirac
order-to-order transitions from fixed-point annihilation. and chiral spin liquids [65, 75, 76] and remains within the
physical space of real couplings for all 𝑁 ≥ 1. The latter,
Kagome quantum magnets. Although the proposed mech- in contrast, collides and annihilates with the infrared stable
anism does not depend on the formation of fractionalized fixed point at a critical value 𝑁c > 2, consistent with previous
quasiparticles, it is possible to emerge also in conjunction with studies [48–50, 83–87]. In the context of the kagome spin
fractionalization phenomena. To illustrate this point, consider model, the resulting runaway flow can be understood as a low-
the extended Heisenberg model on the kagome lattice, defined temperature instability of the U(1) Dirac spin liquid towards
 5

valence bond solid order [88]. This implies that the remaining [11] M. Song, J. Zhao, L. Janssen, M. M. Scherer, and Z. Y. Meng,
quantum critical fixed point for 𝑁 < 𝑁c describes a continu- Deconfined quantum criticality lost, arXiv:2307.02547.
ous transition between two distinct symmetry-broken phases [12] A. Nahum, J. T. Chalker, P. Serna, M. Ortuño, and A. M.
on the kagome lattice, as depicted in Fig. 4(b). Somoza, Deconfined Quantum Criticality, Scaling Violations,
and Classical Loop Models, Phys. Rev. X 5, 041048 (2015).
Conclusions. We have proposed a mechanism for generic [13] J. Takahashi, H. Shao, B. Zhao, W. Guo, and A. W. Sandvik,
continuous order-to-order quantum phase transitions that op- SO(5) multicriticality in two-dimensional quantum magnets,
erates independently of fractionalization. This mechanism is arXiv:2405.06607.
based on the collision and annihilation of two RG fixed points, [14] Z. Zhou, L. Hu, W. Zhu, and Y.-C. He, SO(5) Deconfined Phase
and may be realized in pyrochlore iridates and kagome quan- Transition under the Fuzzy-Sphere Microscope: Approximate
Conformal Symmetry, Pseudo-Criticality, and Operator Spec-
tum magnets. Further possible realizations in quantum im-
trum, Phys. Rev. X 14, 021044 (2024).
purity models and quantum chromodynamics with additional [15] A. A. Abrikosov and S. D. Beneslavskii, Possible existence of
four-fermion interactions are discussed in the SM [29]. A dis- substances intermediate between metals and dielectrics, Zh.
tinctive experimental or numerical signature of the mechanism Eksp. Teor. Fiz. 59, 1280 (1970), [Sov. Phys. JETP 32, 699
is a strong asymmetry in the effective energy scales on either (1971)].
side of the transition. [16] A. A. Abrikosov, Calculation of critical indices for zero-gap
semiconductors, Zh. Eksp. Teor. Fiz. 66, 1443 (1974), [Sov.
Acknowledgments. We thank Igor Boettcher, Rufus Boy-
Phys. JETP 39, 709 (1974)].
ack, Shankar Ganesh, Igor F. Herbut, Joseph Maciejko, [17] E.-G. Moon, C. Xu, Y. B. Kim, and L. Balents, Non-Fermi-
Zi Yang Meng, Max A. Metlitski, Adam Nahum, Shouryya Liquid and Topological States with Strong Spin-Orbit Cou-
Ray, Michael M. Scherer, Carsten Timm, Matthias Vojta, pling, Phys. Rev. Lett. 111, 206401 (2013).
and Manuel Weber for illuminating discussions. This work [18] S. Murakami, N. Nagosa, and S.-C. Zhang, SU(2) non-Abelian
has been supported by the Deutsche Forschungsgemeinschaft holonomy and dissipationless spin current in semiconductors,
(DFG) through SFB 1143 (A07, Project No. 247310070), the Phys. Rev. B 69, 235206 (2004).
[19] W. Witczak-Krempa, G. Chen, Y. B. Kim, and L. Balents, Cor-
Würzburg-Dresden Cluster of Excellence ct.qmat (EXC 2147,
related Quantum Phenomena in the Strong Spin-Orbit Regime,
Project No. 390858490), and the Emmy Noether program Annu. Rev. Condens. Matter Phys. 5, 57 (2014).
(JA2306/4-1, Project No. 411750675). [20] J. M. Luttinger, Quantum Theory of Cyclotron Resonance
in Semiconductors: General Theory, Phys. Rev. 102, 1030
(1956).
[21] I. F. Herbut and L. Janssen, Topological Mott Insulator in
Three-Dimensional Systems with Quadratic Band Touching,
[1] T. Senthil, A. Vishwanath, L. Balents, S. Sachdev, and M. P. A. Phys. Rev. Lett. 113, 106401 (2014).
Fisher, Deconfined Quantum Critical Points, Science 303, 1490 [22] L. Janssen and I. F. Herbut, Nematic quantum criticality in
(2004). three-dimensional Fermi system with quadratic band touching,
[2] T. Senthil, L. Balents, S. Sachdev, A. Vishwanath, and M. P. A. Phys. Rev. B 92, 045117 (2015).
Fisher, Quantum criticality beyond the Landau-Ginzburg- [23] L. Savary, E.-G. Moon, and L. Balents, New Type of Quantum
Wilson paradigm, Phys. Rev. B 70, 144407 (2004). Criticality in the Pyrochlore Iridates, Phys. Rev. X 4, 041027
[3] T. Senthil, Deconfined quantum critical points: a review, (2014).
arXiv:2306.12638. [24] L. Janssen and I. F. Herbut, Excitonic instability of three-
[4] J. Lou, A. W. Sandvik, and N. Kawashima, Antiferromagnetic dimensional gapless semiconductors: Large-𝑁 theory, Phys.
to valence-bond-solid transitions in two-dimensional SU(𝑁) Rev. B 93, 165109 (2016).
Heisenberg models with multispin interactions, Phys. Rev. B [25] L. Janssen and I. F. Herbut, Phase diagram of electronic sys-
80, 180414 (2009). tems with quadratic Fermi nodes in 2 < 𝑑 < 4: 2+𝜖 expansion,
[5] A. Nahum, P. Serna, J. T. Chalker, M. Ortuño, and A. M. 4 − 𝜖 expansion, and functional renormalization group, Phys.
Somoza, Emergent SO(5) Symmetry at the Néel to Valence- Rev. B 95, 075101 (2017).
Bond-Solid Transition, Phys. Rev. Lett. 115, 267203 (2015). [26] I. Boettcher and I. F. Herbut, Anisotropy induces non-
[6] Z. Deng, L. Liu, W. Guo, and H. qing Lin, Diagnosing 𝑆𝑂 (5) Fermi-liquid behavior and nematic magnetic order in three-
Symmetry and First-Order Transition in the 𝐽 − 𝑄 3 Model via dimensional Luttinger semimetals, Phys. Rev. B 95, 075149
Entanglement Entropy, arXiv:2401.12838. (2017).
[7] J. D’Emidio and A. W. Sandvik, Entanglement entropy and [27] D. J. Moser and L. Janssen, Quasiuniversality from all-in–all-
deconfined criticality: emergent SO(5) symmetry and proper out Weyl quantum criticality in pyrochlore iridates, Phys. Rev.
lattice bipartition, arXiv:2401.14396. B 109, L081111 (2024).
[8] N. Ma, G.-Y. Sun, Y.-Z. You, C. Xu, A. Vishwanath, A. W. [28] K. Ladovrechis, T. Meng, and B. Roy, Competing magnetic
Sandvik, and Z. Y. Meng, Dynamical signature of fractional- orders and multipolar Weyl fermions in 227 pyrochlore iridates,
ization at a deconfined quantum critical point, Phys. Rev. B 98, Phys. Rev. B 103, L241116 (2021).
174421 (2018). [29] See Supplemental Material, which additionally includes
[9] C. Wang, A. Nahum, M. A. Metlitski, C. Xu, and T. Senthil, Refs. [87–115], for (i) the Luttinger model in the Hubbard-
Deconfined Quantum Critical Points: Symmetries and Duali- Stratonovich formulation, (ii) definitions and numerical values
ties, Phys. Rev. X 7, 031051 (2017). of solid-angle integrals, (iii) technical details of the mean-field
[10] J. Zhao, Y.-C. Wang, Z. Yan, M. Cheng, and Z. Y. Meng, Scal- analysis, (iv) the derivation of the flow equations, (v) a discus-
ing of Entanglement Entropy at Deconfined Quantum Critical- sion of the fixed-point structure, (vi) details on obtaining the
ity, Phys. Rev. Lett. 128, 010601 (2022). phase diagrams by integrating out the flow equations, (vii) a
 6

discussion of higher-loop corrections, (viii) a discussion of the [52] H. Gies and J. Jaeckel, Chiral phase structure of QCD with
RG flow in the limit of infinite order-parameter masses, as well many flavors, Eur. Phys. J. C 46, 433 (2006).
as (ix) further examples for continuous order-to-order transi- [53] J. Braun and H. Gies, Scaling laws near the conformal window
tions from fixed-point annihilation in the context of kagome of many-flavor QCD, J. High Energy Phys. 05 (2010) 060.
quantum magnets, quantum impurity models, and quantum [54] J. Braun, C. S. Fischer, and H. Gies, Beyond Miransky scaling,
chromodynamics with additional four-fermion interactions. Phys. Rev. D 84, 034045 (2011).
[30] H. Gies and C. Wetterich, Renormalization flow of bound [55] Y. Kusafuka and H. Terao, Fixed point merger in the 𝑆𝑈 (𝑁)
states, Phys. Rev. D 65, 065001 (2002). gauge beta functions, Phys. Rev. D 84, 125006 (2011).
[31] J. M. Pawlowski, Aspects of the functional renormalisation [56] F. Kuipers, U. Gürsoy, and Y. Kuznetsov, Bifurcations in the
group, Ann. Phys. 322, 2831 (2007). RG-flow of QCD, J. High Energy Phys. 07 (2019) 075.
[32] S. Floerchinger and C. Wetterich, Exact flow equation for com- [57] M. R. Norman, Colloquium: Herbertsmithite and the search for
posite operators, Phys. Lett. B 680, 371 (2009). the quantum spin liquid, Rev. Mod. Phys. 88, 041002 (2016).
[33] D. B. Kaplan, J.-W. Lee, D. T. Son, and M. A. Stephanov, [58] B. Fåk, E. Kermarrec, L. Messio, B. Bernu, C. Lhuillier,
Conformality lost, Phys. Rev. D 80, 125005 (2009). F. Bert, P. Mendels, B. Koteswararao, F. Bouquet, J. Ollivier,
[34] V. Gorbenko, S. Rychkov, and B. Zan, Walking, weak first- A. D. Hillier, A. Amato, R. H. Colman, and A. S. Wills, Kapel-
order transitions, and complex CFTs, J. High Energy Phys. 10 lasite: A Kagome Quantum Spin Liquid with Competing In-
(2018) 108. teractions, Phys. Rev. Lett. 109, 037208 (2012).
[35] A. Nahum, Fixed point annihilation for a spin in a fluctuating [59] X. Hong, M. Behnami, L. Yuan, B. Li, W. Brenig, B. Büchner,
field, Phys. Rev. B 106, L081109 (2022). Y. Li, and C. Hess, Heat transport of the kagome Heisenberg
[36] H. Hu and Q. Si, Kondo destruction and fixed-point annihila- quantum spin liquid candidate YCu3 (OH) 6.5 Br2.5 : Localized
tion in a Bose-Fermi Kondo model, arXiv:2207.08744. magnetic excitations and a putative spin gap, Phys. Rev. B 106,
[37] M. Weber and M. Vojta, SU(2)-Symmetric Spin-Boson Model: L220406 (2022).
Quantum Criticality, Fixed-Point Annihilation, and Duality, [60] S. Suetsugu, T. Asaba, Y. Kasahara, Y. Kohsaka, K. Totsuka,
Phys. Rev. Lett. 130, 186701 (2023). B. Li, Y. Zhao, Y. Li, M. Tokunaga, and Y. Matsuda, Emer-
[38] M. Weber, Tunable quantum criticality and pseudocriticality gent Spin-Gapped Magnetization Plateaus in a Spin-1/2 Per-
across the fixed-point annihilation in the anisotropic spin- fect Kagome Antiferromagnet, Phys. Rev. Lett. 132, 226701
boson model, arXiv:2403.02400. (2024).
[39] V. Gorbenko, S. Rychkov, and B. Zan, Walking, Weak first- [61] Y. Ran, M. Hermele, P. A. Lee, and X.-G. Wen, Projected-
order transitions, and Complex CFTs II. Two-dimensional Potts Wave-Function Study of the Spin-1/2 Heisenberg Model on
model at 𝑄 > 4, SciPost Phys. 5, 050 (2018). the Kagomé Lattice, Phys. Rev. Lett. 98, 117205 (2007).
[40] H. Ma and Y.-C. He, Shadow of complex fixed point: Approx- [62] S. Yan, D. A. Huse, and S. R. White, Spin-Liquid Ground State
imate conformality of 𝑄 > 4 Potts model, Phys. Rev. B 99, of the 𝑆 = 1/2 Kagome Heisenberg Antiferromagnet, Science
195130 (2019). 332, 1173 (2011).
[41] J. L. Jacobsen and K. J. Wiese, Lattice Realization of Complex [63] S. Depenbrock, I. P. McCulloch, and U. Schollwöck, Nature
Conformal Field Theories: Two-Dimensional Potts Model with of the Spin-Liquid Ground State of the 𝑆 = 1/2 Heisenberg
𝑄 > 4 States, Phys. Rev. Lett. 133, 077101 (2024). Model on the Kagome Lattice, Phys. Rev. Lett. 109, 067201
[42] Y. Tang, H. Ma, Q. Tang, Y.-C. He, and W. Zhu, Reclaiming the (2012).
Lost Conformality in a non-Hermitian Quantum 5-state Potts [64] Y. Iqbal, D. Poilblanc, and F. Becca, Vanishing spin gap in a
Model, arXiv:2403.00852. competing spin-liquid phase in the kagome Heisenberg anti-
[43] B. I. Halperin, T. C. Lubensky, and S.-k. Ma, First-Order Phase ferromagnet, Phys. Rev. B 89, 020407 (2014).
Transitions in Superconductors and Smectic-𝐴 Liquid Crystals, [65] Y.-C. He, M. P. Zaletel, M. Oshikawa, and F. Pollmann, Sig-
Phys. Rev. Lett. 32, 292 (1974). natures of Dirac Cones in a DMRG Study of the Kagome
[44] B. Ihrig, N. Zerf, P. Marquard, I. F. Herbut, and M. M. Scherer, Heisenberg Model, Phys. Rev. X 7, 031020 (2017).
Abelian Higgs model at four loops, fixed-point collision, and [66] H. J. Liao, Z. Y. Xie, J. Chen, Z. Y. Liu, H. D. Xie, R. Z. Huang,
deconfined criticality, Phys. Rev. B 100, 134507 (2019). B. Normand, and T. Xiang, Gapless Spin-Liquid Ground State
[45] R. Ma and C. Wang, Theory of deconfined pseudocriticality, in the 𝑆 = 1/2 Kagome Antiferromagnet, Phys. Rev. Lett. 118,
Phys. Rev. B 102, 020407 (2020). 137202 (2017).
[46] A. Nahum, Note on Wess-Zumino-Witten models and quasi- [67] X. Chen, S.-J. Ran, T. Liu, C. Peng, Y.-Z. Huang, and G. Su,
universality in 2 + 1 dimensions, Phys. Rev. B 102, 201116 Thermodynamics of spin-1/2 Kagomé Heisenberg antiferro-
(2020). magnet: algebraic paramagnetic liquid and finite-temperature
[47] B. Hawashin, A. Eichhorn, L. Janssen, M. M. Scherer, and phase diagram, Sci. Bull. 63, 1545 (2018).
S. Ray, The Nordic-walking mechanism and its explanation of [68] J. Schnack, J. Schulenburg, and J. Richter, Magnetism of the
deconfined pseudocriticality from Wess-Zumino-Witten the- 𝑁 = 42 kagome lattice antiferromagnet, Phys. Rev. B 98,
ory, arXiv:2312.11614. 094423 (2018).
[48] J. Braun, H. Gies, L. Janssen, and D. Roscher, Phase structure [69] A. M. Läuchli, J. Sudan, and R. Moessner, 𝑆 = 21 kagome
of many-flavor QED3 , Phys. Rev. D 90, 036002 (2014). Heisenberg antiferromagnet revisited, Phys. Rev. B 100,
[49] L. Janssen, Spontaneous breaking of Lorentz symmetry in (2 + 155142 (2019).
𝜖)-dimensional QED, Phys. Rev. D 94, 094013 (2016). [70] Y.-C. He and Y. Chen, Distinct Spin Liquids and Their Transi-
[50] I. F. Herbut, Chiral symmetry breaking in three-dimensional tions in Spin-1/2 𝑋 𝑋 𝑍 Kagome Antiferromagnets, Phys. Rev.
quantum electrodynamics as fixed point annihilation, Phys. Lett. 114, 037201 (2015).
Rev. D 94, 025036 (2016). [71] Y.-C. He, S. Bhattacharjee, F. Pollmann, and R. Moessner,
[51] S. Gukov, RG flows and bifurcations, Nucl. Phys. B 919, 583 Kagome Chiral Spin Liquid as a Gauged 𝑈 (1) Symmetry Pro-
(2017). tected Topological Phase, Phys. Rev. Lett. 115, 267209 (2015).
 7

[72] W.-J. Hu, W. Zhu, Y. Zhang, S. Gong, F. Becca, and D. N. Berlin, Heidelberg, 2007).
Sheng, Variational Monte Carlo study of a chiral spin liquid in [92] L. Janssen and H. Gies, Critical behavior of the (2 + 1)-
the extended Heisenberg model on the kagome lattice, Phys. dimensional Thirring model, Phys. Rev. D 86, 105007 (2012).
Rev. B 91, 041124 (2015). [93] I. F. Herbut and Z. Tešanović, Critical Fluctuations in Super-
[73] S.-S. Gong, W. Zhu, L. Balents, and D. N. Sheng, Global phase conductors and the Magnetic Field Penetration Depth, Phys.
diagram of competing ordered and quantum spin-liquid phases Rev. Lett. 76, 4588 (1996).
on the kagome lattice, Phys. Rev. B 91, 075112 (2015). [94] I. F. Herbut and Z. Tešanović, Herbut and Tešanović Reply:,
[74] A. Wietek, A. Sterdyniak, and A. M. Läuchli, Nature of chiral Phys. Rev. Lett. 78, 980 (1997).
spin liquids on the kagome lattice, Phys. Rev. B 92, 125122 [95] M. S. Block, R. G. Melko, and R. K. Kaul, Fate of CP 𝑁 −1
(2015). Fixed Points with 𝑞 Monopoles, Phys. Rev. Lett. 111, 137202
[75] Y.-C. He, Y. Fuji, and S. Bhattacharjee, Kagome spin liquid: (2013).
a deconfined critical phase driven by 𝑈 (1) gauge fluctuation, [96] L. Fei, S. Giombi, and I. R. Klebanov, Critical 𝑂 (𝑁) models
arXiv:1512.05381. in 6 − 𝜖 dimensions, Phys. Rev. D 90, 025018 (2014).
[76] L. Janssen and Y.-C. He, Critical behavior of the QED3 -Gross- [97] I. F. Herbut and L. Janssen, Critical 𝑂 (2) and 𝑂 (3) 𝜑4 theories
Neveu model: Duality and deconfined criticality, Phys. Rev. B near six dimensions, Phys. Rev. D 93, 085005 (2016).
96, 205113 (2017). [98] J. A. Gracey, I. F. Herbut, and D. Roscher, Tensor 𝑂 (𝑁) model
[77] B. Ihrig, L. Janssen, L. N. Mihaila, and M. M. Scherer, Decon- near six dimensions: Fixed points and conformal windows
fined criticality from the QED3 -Gross-Neveu model at three from four loops, Phys. Rev. D 98, 096014 (2018).
loops, Phys. Rev. B 98, 115163 (2018). [99] L. Rosa, P. Vitale, and C. Wetterich, Critical Exponents of the
[78] N. Zerf, P. Marquard, R. Boyack, and J. Maciejko, Critical be- Gross-Neveu Model from the Effective Average Action, Phys.
havior of the QED3 -Gross-Neveu-Yukawa model at four loops, Rev. Lett. 86, 958 (2001).
Phys. Rev. B 98, 165125 (2018). [100] N. Dupuis, L. Canet, A. Eichhorn, W. Metzner, J. Pawlowski,
[79] R. Boyack, A. Rayyan, and J. Maciejko, Deconfined criticality M. Tissier, and N. Wschebor, The nonperturbative functional
in the QED3 Gross-Neveu-Yukawa model: The 1/𝑁 expansion renormalization group and its applications, Phys. Rep. 910, 1
revisited, Phys. Rev. B 99, 195135 (2019). (2021).
[80] E. Dupuis, R. Boyack, and W. Witczak-Krempa, Anomalous [101] Y. Huh and S. Sachdev, Renormalization group theory of ne-
Dimensions of Monopole Operators at the Transitions between matic ordering in 𝑑-wave superconductors, Phys. Rev. B 78,
Dirac and Topological Spin Liquids, Phys. Rev. X 12, 031012 064512 (2008).
(2022). [102] J. Schwab, L. Janssen, K. Sun, Z. Y. Meng, I. F. Herbut,
[81] N. Karthik and R. Narayanan, No evidence for bilinear conden- M. Vojta, and F. F. Assaad, Nematic Quantum Criticality in
sate in parity-invariant three-dimensional QED with massless Dirac Systems, Phys. Rev. Lett. 128, 157203 (2022).
fermions, Phys. Rev. D 93, 045020 (2016). [103] H. Gies and L. Janssen, UV fixed-point structure of the three-
[82] N. Karthik and R. Narayanan, Scale invariance of parity- dimensional Thirring model, Phys. Rev. D 82, 085018 (2010).
invariant three-dimensional QED, Phys. Rev. D 94, 065026 [104] V. P. Gusynin and P. K. Pyatkovskiy, Critical number of
(2016). fermions in three-dimensional QED, Phys. Rev. D 94, 125009
[83] T. Appelquist, D. Nash, and L. C. R. Wijewardhana, Critical (2016).
Behavior in (2+1)-Dimensional QED, Phys. Rev. Lett. 60, 2575 [105] A. V. Kotikov and S. Teber, Critical behavior of (2 + 1)-
(1988). dimensional QED: 1/𝑁 𝑓 corrections in an arbitrary nonlocal
[84] K.-I. Kubota and H. Terao, Dynamical Symmetry Breaking in gauge, Phys. Rev. D 94, 114011 (2016).
QED3 from the Wilson RG Point of View, Progr. Theor. Phys. [106] A. M. Sengupta, Spin in a fluctuating field: The Bose(+Fermi)
105, 809 (2001). Kondo models, Phys. Rev. B 61, 4041 (2000).
[85] K. Kaveh and I. F. Herbut, Chiral symmetry breaking in three- [107] S. Sachdev, C. Buragohain, and M. Vojta, Quantum Impurity in
dimensional quantum electrodynamics in the presence of irrel- a Nearly Critical Two-Dimensional Antiferromagnet, Science
evant interactions: A renormalization group study, Phys. Rev. 286, 2479 (1999).
B 71, 184519 (2005). [108] S. Sachdev, Quantum impurity in a magnetic environment, J.
[86] T. Grover, Entanglement Monotonicity and the Stability of Stat. Phys. 115, 47 (2004).
Gauge Theories in Three Spacetime Dimensions, Phys. Rev. [109] M. Vojta, Impurity quantum phase transitions, Philos. Mag.
Lett. 112, 151601 (2014). 86, 1807 (2006).
[87] L. Di Pietro, Z. Komargodski, I. Shamir, and E. Stamou, Quan- [110] L. Zhu and Q. Si, Critical local-moment fluctuations in the
tum Electrodynamics in 𝑑 = 3 from the 𝜖 Expansion, Phys. Rev. Bose-Fermi Kondo model, Phys. Rev. B 66, 024426 (2002).
Lett. 116, 131601 (2016). [111] G. Zaránd and E. Demler, Quantum phase transitions in the
[88] M. Hermele, T. Senthil, and M. P. A. Fisher, Algebraic spin Bose-Fermi Kondo model, Phys. Rev. B 66, 024427 (2002).
liquid as the mother of many competing orders, Phys. Rev. B [112] G. Cuomo, Z. Komargodski, M. Mezei, and A. Raviv-Moshe,
72, 104404 (2005). Spin impurities, Wilson lines and semiclassics, J. High Energy
[89] S. Ray and L. Janssen, Gross-Neveu-Heisenberg criticality Phys. 06 (2022) 112.
from competing nematic and antiferromagnetic orders in bi- [113] H. Gies, J. Jaeckel, and C. Wetterich, Towards a renormalizable
layer graphene, Phys. Rev. B 104, 045101 (2021). standard model without a fundamental Higgs scalar, Phys. Rev.
[90] M. Vojta, Y. Zhang, and S. Sachdev, Renormalization group D 69, 105008 (2004).
analysis of quantum critical points in 𝑑-wave superconductors, [114] W. E. Caswell, Asymptotic Behavior of Non-Abelian Gauge
Int. J. Mod. Phys. B 14, 3719 (2000). Theories to Two-Loop Order, Phys. Rev. Lett. 33, 244 (1974).
[91] M. S. Dresselhaus, G. Dresselhaus, and A. Jorio, Group the- [115] T. Banks and A. Zaks, On the phase structure of vector-like
ory: application to the physics of condensed matter (Springer gauge theories with massless fermions, Nucl. Phys. B 196, 189
(1982).
 Supplemental Material for
“Continuous order-to-order quantum phase transitions from fixed-point annihilation”
David J. Moser and Lukas Janssen
Institut für Theoretische Physik and Würzburg-Dresden Cluster of Excellence ct.qmat, TU Dresden, 01062 Dresden, Germany
(Dated: December 11, 2024)
The Supplemental Material contains (i) the Luttinger model in the Hubbard-Stratonovich formulation, (ii) def-
initions and numerical values of solid-angle integrals, (iii) technical details of the mean-field analysis, (iv) the
derivation of the flow equations, (v) a discussion of the fixed-point structure, (vi) details on obtaining the phase
diagrams by integrating out the flow equations, (vii) a discussion of higher-loop corrections, (viii) a discussion
of the RG flow in the limit of infinite order-parameter masses, as well as (ix) further examples for continuous
order-to-order transitions from fixed-point annihilation in the context of kagome quantum magnets, quantum
impurity models, and quantum chromodynamics with additional four-fermion interactions.
arXiv:2412.06890v1 [cond-mat.str-el] 9 Dec 2024

I. HUBBARD-STRATONOVICH FORMULATION The fermion-boson model defined by the Euclidean action

In this supplemental section, we show how the interactions 𝑆FB = 𝑆0 + 𝑆 𝑎 + 𝑆 𝜙 + 𝑆 𝜑 (1.4)

in the Luttinger model can be rewritten in terms of Hubbard-
Stratonovich fields 𝑎, 𝜙, and 𝜑. The Hubbard-Stratonovich is quadratic in the Hubbard-Stratonovich fields 𝑎, 𝜙, and 𝜑.
formulation of the model is useful for the mean-field and renor- The latter can therefore, in principle, be integrated out. The
malization group (RG) analyses. In particular, this formulation original microscopic, purely fermionic action 𝑆 = 𝑆0 + 𝑆int
is crucial in order to explicitly compute order parameters using presented in the main text can then be recovered from the above
the RG, since it avoids the runaway flows encountered in the fermion-boson action in the limit of infinite order-parameter
2 → ∞, with fixed ratios
masses 𝑟 1,2 → ∞ and 𝑔1,2
purely fermionic formulations [1–3].
We introduce the scalar Coulomb field 𝑎 via
∫   𝑔12 𝑔22
3 1 2 i𝑒 † 𝐺1 = and 𝐺2 = , (1.5)
𝑆 𝑎 = d𝜏 d 𝑥® (∇𝑎) + √ 𝑎𝜓𝑖 𝜓𝑖 , (1.1) 𝑟1 𝑟2
2 𝑁
and with 𝑐 = 𝜆 = 0.
where the summation over repeated indices 𝑖 = 1, . . . , 𝑁 has
been implicitly assumed. All-in-all-out (AIAO) fluctuations
are parametrized by an Ising order-parameter field 𝜙, which
II. SOLID-ANGLE INTEGRALS
couples to the electronic quasiparticles as [3–6]
∫  
1 𝑔1 In this supplemental section, we provide definitions and
𝑆 𝜙 = d𝜏 d3 𝑥® 𝜙(𝑟 1 − ∇2 )𝜙 + √ 𝜙𝜓𝑖† 𝛾45 𝜓𝑖 , (1.2)
2 𝑁 numerical values of the dimensionless functions 𝑓𝑖 (𝛿), with
𝑖 ∈ {1, 1t, 1e, 2, 2t, 2e, 3, 3t, 3e, 3t, 4, 4t, 4e, 4tt, 4tt′ , 4ee, 4et,
where 𝑟 1 represents the tuning parameter for the AIAO tran- 4tt, 4et, 𝑒 2 }, and 𝑓𝑔2 (𝛿) as function of the anisotropy parameter
sition and 𝑔1 is the AIAO Yukawa coupling. Nematic fluc- 𝛿, occurring in the mean-field calculation (see Sec. III) and in
tuations are parametrized by a five-component tensor order- the RG analysis (see Secs. IV–VIII). These are bounded and
parameter field 𝜑 𝑎 , which couples to the electronic quasipar- continuous functions of order unity with 𝑓𝑖 > 0 ( 𝑓𝑔2 ≥ −2/3)
ticles as [2, 7, 8] for 𝛿 ∈ [−1, 1] and 𝑓𝑖 = 1 ( 𝑓𝑔2 = 0) for 𝛿 = 0. Some of these
∫ " 5 have already been defined in Ref. [3], reading
1 ∑︁ ∫
𝑆 𝜑 = d𝜏 d3 𝑥® 𝜑 𝑎 (𝑟 2 − 𝑐𝜕𝜏2 − ∇2 )𝜑 𝑎 1 1
2 𝑎=1 𝑓1 (𝛿) B dΩ 1/2 , (2.1)
∑︁ ∑︁  4𝜋 𝑋˜
5 5
∫
+√
𝑔2
𝜑 𝑎 𝜓𝑖† 𝛾 𝑎 𝜓𝑖 + √
𝜆
tr(Λ𝑎 Λ𝑏 Λ𝑐 )𝜑 𝑎 𝜑 𝑏 𝜑 𝑐 , 5 𝑑˜12
𝑁 𝑎=1 𝑁 𝑎,𝑏,𝑐=1 𝑓 1t (𝛿) B dΩ , (2.2)
4𝜋 𝑋˜ 1/2
(1.3) ∫
5 𝑑˜42
𝑓1e (𝛿) B dΩ 1/2 , (2.3)
where 𝑟 2 represents the tuning parameter for the nematic tran- 4𝜋 𝑋˜ ∫
sition, 𝑔2 is the nematic Yukawa coupling, and 𝜆 parametrizes 1 1
𝑓2 (𝛿) B (1 − 𝛿) (1 + 𝛿) dΩ 3/2 , (2.4)
the bosonic self-interaction. Λ𝑎 denote the five real Gell- 4𝜋 ˜
𝑋
Mann matrices in three spatial dimensions [9]. The parameter ∫
5 𝑑˜12
𝑐 is power-counting irrelevant and included for regularization 𝑓2t (𝛿) B (1 + 𝛿) dΩ 3/2 , (2.5)
purposes. Quartic bosonic self-interactions are also power- 4𝜋 𝑋˜
∫
counting irrelevant and are therefore set to zero from the be- 5 𝑑˜42
ginning. 𝑓 2e (𝛿) B (1 − 𝛿) dΩ , (2.6)
4𝜋 𝑋˜ 3/2
 2

TABLE I: Values of solid-angle integrals 𝑓𝑖 (𝛿) for limiting cases 𝛿 = ±1 and isotropic case 𝛿 = 0 from, whenever possible, analytical
integration, otherwise determined numerically.

𝑓𝑖 𝑓𝑖 (−1) 𝑓𝑖 (0) 𝑓𝑖 (+1) 𝑓𝑖 𝑓𝑖 (−1) 𝑓𝑖 (0) 𝑓𝑖 (+1) 𝑓𝑖 𝑓𝑖 (−1) 𝑓𝑖 (0) 𝑓𝑖 (+1)

𝑓1 1.0942 1 0.8130 𝑓3t 5/9 1 5/9 𝑓4tt′ 7/9 1 7/18
𝑓1t 1.3294 1 0.6225 𝑓3e 5/3 1 5/12 𝑓4ee 7/3 1 7/16
𝑓1e 0.7415 1 1.0987 𝑓3t 35/54 1 0.7262 𝑓4et 7/9 1 7/24
𝑓2 1 1 1/2 𝑓4 16/5 1 8/5 𝑓4tt 77/162 1 0.8876
𝑓2t 5/6 1 0.6775 𝑓4t 4/3 1 4/9 𝑓4et 77/108 1 77/144
𝑓2e 1.3678 1 5/8 𝑓4e 4/3 1 1 𝑓𝑒2 16/9 1 4/3
𝑓3 4/3 1 2/3 𝑓4tt 7/27 1 7/9 𝑓𝑔2 16/9 0 −2/3

∫
1 3 3 1 For the Coulomb and order-parameter anomalous dimensions,
𝑓3 (𝛿) B (1 − 𝛿) (1 + 𝛿) dΩ 5/2 , (2.7)
4𝜋 𝑋˜ it is convenient to define two further functions as combinations
∫ ˜2 of the above-defined ones,
5 𝑑 1
(1 − 𝛿) (1 + 𝛿) 3 dΩ 5/2 (2.8)
1 
𝑓3t (𝛿) B ,
4𝜋 𝑋˜
∫ 𝑓𝑒2 (𝛿) B 2(1 − 𝛿) 2 + 3(1 + 𝛿) 2 𝑓2 (𝛿)
5 𝑑˜42 3 
𝑓3e (𝛿) B (1 − 𝛿) 3 (1 + 𝛿) dΩ 5/2 , (2.9) 2 2 3
4𝜋 𝑋˜ (1 − 𝛿) 2 𝑓3e (𝛿) + (1 + 𝛿) 2 𝑓3t (𝛿)
∫ −
3 5 5
35 𝑑˜1 𝑑˜2 𝑑˜3 
𝑓3t (𝛿) B √ (1 + 𝛿) 3 dΩ , (2.10) 12 𝛿2 36 𝛿2 (1 − 𝛿)
34𝜋 𝑋˜ 5/2 − 𝑓 3t (𝛿) + 𝑓 (𝛿) ,
5 (1 + 𝛿) 2 35 (1 + 𝛿) 2 3t
∫ ∫𝜋 ∫ 2𝜋 (2.20)
where dΩ = 0 d𝜃 sin 𝜃 0 d𝜙 denotes the integration over
Í 𝑞2 𝑞
2 2(1 − 𝛿) 2
− (1 + 𝛿) 2
the solid angle, 𝑋˜ (𝜃, 𝜙) = (1 − 𝛿) 2 + 12𝛿 𝑖< 𝑗 𝑞𝑖2 𝑞2𝑗 , and 𝑓𝑔2 (𝛿) B
4
𝑓2 (𝛿)
𝑑˜𝑎 (𝜃, 𝜙) B 𝑑 𝑎 ( 𝑞)/𝑞
® 2 are the ℓ = 2 real spherical harmon- 8𝛿2 − 6(1 − 𝛿) 2 (1 + 𝛿) 2 − 11(1 + 𝛿) 4
ics. Upon spatial rotations, the latter transform under the irre- + 𝑓3t (𝛿)
20(1 + 𝛿) 2
ducible representation T2g (Eg ) of the octahedral point group
Oh for 𝑎 = 1, 2, 3 (𝑎 = 4, 5), and the indices t and e of the 4(1 − 𝛿) 2 + 9(1 + 𝛿) 2 6𝛿2 (1 − 𝛿)
+ 𝑓3e (𝛿) − 𝑓 (𝛿)
functions 𝑓𝑖 indicate the type of spherical harmonics involved 30 35(1 + 𝛿) 2 3t
(  
in the integral. In addition to the above-defined functions, the (1 − 𝛿) 2 −4(1 − 𝛿) 2 + 7(1 + 𝛿) 2
loop expansion of the AIAO order-parameter field theory [6] +
14(1 + 𝛿) 2
gave rise to the following new solid-angle integrals,  )
2𝛿2 −(1 − 𝛿) 2 + 9(1 + 𝛿) 2
∫ + 𝑓4tt (𝛿)
1 1 7(1 + 𝛿) 4
𝑓4 (𝛿) B (1 − 𝛿) 5 (1 + 𝛿) 5 dΩ (2.11)
(
,
4𝜋 ˜
𝑋 7/2  
∫ (1 − 𝛿) 2 −5(1 − 𝛿) 2 + 11(1 + 𝛿) 2
5 𝑑˜12 +
𝑓4t (𝛿) B (1 − 𝛿) 3 (1 + 𝛿) 5 dΩ 7/2 , (2.12) 42(1 + 𝛿) 2
4𝜋 𝑋˜  )
∫
5 𝑑˜42 2𝛿2 (1 − 𝛿) 2 + 18(1 + 𝛿) 2
𝑓4e (𝛿) B (1 − 𝛿) 5 (1 + 𝛿) 3 dΩ 7/2 , (2.13) + 𝑓4tt′ (𝛿)
4𝜋 𝑋˜ 21(1 + 𝛿) 4
∫
35 𝑑˜2 · 𝑑˜2 4(1 − 𝛿) 2 54𝛿2 (1 − 𝛿)
𝑓4tt (𝛿) B (1 − 𝛿)(1 + 𝛿) 5
dΩ 1 7/21 , (2.14) − 𝑓4ee (𝛿) + 𝑓 (𝛿)
12𝜋 𝑋˜ 21 77(1 + 𝛿) 2 4tt
∫  
35 𝑑 · 𝑑˜2
˜2
4𝛿 (1 − 𝛿) 2 + 3(1 + 𝛿) 2
𝑓4tt′ (𝛿) B (1 − 𝛿)(1 + 𝛿) 5 dΩ 1 7/22 , (2.15) − 𝑓4et (𝛿)
4𝜋 𝑋˜ 21(1 + 𝛿) 2
∫
35 𝑑˜2 · 𝑑˜2 12𝛿2 (1 − 𝛿) 2
𝑓4ee (𝛿) B (1 − 𝛿) 5 (1 + 𝛿) dΩ 4 7/24 , (2.16) − 𝑓 (𝛿). (2.21)
12𝜋 𝑋˜ 77(1 + 𝛿) 2 4et
∫
35 𝑑 2 · 𝑑˜2
˜
We emphasize that all 𝑓𝑖 are bounded from above and below
𝑓4et (𝛿) B (1 − 𝛿) 3 (1 + 𝛿) 3 dΩ 1 7/24 , (2.17)
4𝜋 𝑋˜ for all 𝛿 ∈ [−1, 1]; in particular, they remain finite in the
∫
385 𝑑 · 𝑑1 𝑑˜2 𝑑˜3
˜2 ˜ limiting cases 𝛿 = ±1, see Table I. For graphs of the above
𝑓4tt (𝛿) B √ (1 + 𝛿) 5 dΩ 1 7/2 , (2.18) defined functions 𝑓𝑖 , we refer the reader to Refs. [3, 6]. Further,
312𝜋 𝑋˜
∫
𝑓 2 ( 𝛿)

𝑑˜2 · 𝑑˜1 𝑑˜2 𝑑˜3 the two ratios, − 𝑔 𝛿 and 𝑓1e ( 𝛿 ) −𝛿 𝑓1t ( 𝛿 ) , arise frequently in
385
𝑓4et (𝛿) B √ (1 − 𝛿)(1 + 𝛿) 3
dΩ 4 7/2 . (2.19) the subsequent RG analysis (see Sec. IV). Their graphs are
34𝜋 𝑋˜ displayed in Fig. S1. Most importantly, both ratios are well
 3
v
u
t
∑︁  1/2
3

®  .
𝑔2 𝜑 |𝑔1 𝜙|
+ 2(1 − 𝛿)𝑑5 √ ± 2(1 + 𝛿) √ 𝑑 𝑎2 ( 𝑝)
𝑁 𝑁 
𝑎=1

(3.2)

In Eq. (3.1), the first two terms, proportional to 𝑟 1 and 𝑟 2 ,

penalize finite order parameters, while the integral can lower
the energy by causing partial or full band gap opening.
𝑓 2 𝑓 −𝑓
In order to expand the mean-field energy around small val-
FIG. S1: Graphs of two useful expression, − 𝑔𝛿 (left) and 1e 𝛿 1t ues of the order parameters, we introduce polar√coordinates in
(right), arising frequently in the subsequent RG analysis, as function √
field space as 𝜙 = ( 𝑁/𝑔1 ) 𝜒 cos 𝛼 and 𝜑 = ( 𝑁/𝑔2 ) 𝜒 sin 𝛼,
of anisotropy parameter 𝛿. Most importantly, both functions only
with radial part 𝜒 > 0 and angle part 𝛼 ∈ [0, 2𝜋). A fi-
possess a removable singularity at 𝛿 = 0.
nite 𝜒 signals an instability of the Luttinger semimetal state,
while the minimizing angle determines whether AIAO order
𝑓 𝑔2 ( 𝛿 ) (𝛼 ∈ {0, 𝜋}), nematic order (𝛼 ∈ {𝜋/2, 3𝜋/2}), or a coexis-
defined in the isotropic limit 𝛿 → 0, namely lim 𝛿→0 −𝛿 = tence of both orders is realized (𝛼 ∉ {0, 𝜋/2, 𝜋, 3𝜋/2}). Fol-
48 𝑓1e ( 𝛿 ) − 𝑓1t ( 𝛿 )
= 74 . lowing Refs. [6, 7], it is convenient to add two suitably written
35 and lim 𝛿→0 𝛿
zeros to the mean-field energy, corresponding to the constant
and quadratic terms in 𝐸 MF with respect to 𝜒, namely
III. MEAN-FIELD THEORY ∫  ∫ 𝑋˜ 1/4 Λ  
dΩ  1 𝜒 1/2 4 𝑐± 𝑥 2 1 𝑐2±
𝜒5/2
(2𝜋) 3  𝑋˜ 3/4 0
0= d𝑥 𝑥 + + −
In this supplemental section, we provide details of the 2 𝑋˜ 1/2 2 8 𝑋˜

mean-field analysis. Instead of integrating out the Hubbard-
 1/2 5 
Stratonovich fields in the fermion-boson action, Eq. (1.4), one 𝑋˜ Λ 𝑐 ± Λ3 𝜒 Λ𝜒2 𝑐2± Λ𝜒2 
8 𝑋˜ 3/2 
can alternatively integrate out the fermion fields 𝜓𝑖† and 𝜓𝑖 , − + + − , (3.3)
5 6 𝑋˜ 1/2 2 𝑋˜ 1/2
𝑖 = 1, . . . , 𝑁. This leads to an effective action for the Hubbard- 
Stratonovich fields, which is solved exactly in the 𝑁 → ∞ limit √︃
using the saddle-point technique. Ultimately, this amounts to Í3 ˜2
where 𝑐 ± B 2(1 − 𝛿) 𝑑˜5 sin 𝛼 ± 2(1 + 𝛿)|cos 𝛼| 𝑎=1 𝑑 𝑎 with
minimizing the mean-field energy [6, 7, 10], ˜ ˜
∫𝑑 𝑎 ≡ 𝑑 𝑎 (𝜃, 𝜙) the real spherical harmonics defined in Sec. II,
𝑟1 2 𝑟2 2 dΩ the integration over the solid angle in momentum space,
𝐸 MF (𝜙, 𝜑) =
2
𝜙 + 𝜑
2 Í 𝑝2 𝑝
2

∫ Λ 3 h and we abbreviated 𝑋˜ (𝜃, 𝜙) B (1 − 𝛿) 2 + 12𝛿 𝑖< 𝑗 𝑝𝑖2 𝑝2𝑗 .

d 𝑝® i
+𝑁 𝜀 +
( ®
𝑝) + 𝜀 −
( ®
𝑝) (3.1) Adding those zeros allows us to express the mean-field energy
0 (2𝜋)
3 𝜙, 𝜑 𝜙, 𝜑
for small 𝜒 ≪ Λ2 as
!
with respect to the order-parameter fields 𝜙 and 𝜑. Here, 𝐸 MF ( 𝜒, 𝛼) 1 𝑟 1 cos2 𝛼 𝑟 2 sin2 𝛼  𝜒 2
we used the fact that minimization of the mean-field energy = + − 𝐼2 (𝛼, 𝛿)
𝑁Λ 5 2 2
𝑔1 Λ 2
𝑔2 Λ Λ2
implies the Coulomb field vanishes at the mean-field level,
 𝜒  5/2   
and assumed that the nematic order parameter takes the form 𝜒 3
+ 𝐼5/2 (𝛼, 𝛿) 2 +O , (3.4)
(𝜑 𝑎 ) = (0, 0, 0, 0, 𝜑). This form implies that the nematic Λ Λ2
phase is uniaxial, characterized by a single director with a
residual discrete rotational symmetry remaining intact in the where we have subtracted an additive constant, which is irrele-
plane perpendicular to the director, and is believed to realize vant for the minimization problem, and introduced the dimen-
the smallest mean-field energy among the different possible sionless integrals
nematic states [7]. Λ is the ultraviolet cutoff. In order to allow ∫  2 
for an easier comparison with numerics and/or experiment, dΩ 𝑐 + + 𝑐2− 2
𝐼2 (𝛼, 𝛿) B − + 1/2 (3.5)
we work with dimensionful quantities 𝐸 MF , 𝑟 1 , 𝑟 2 , 𝑔1 , 𝑔2 , (2𝜋) 3 4 𝑋˜ 3/2 𝑋˜
𝜙, and 𝜑 in this section. Their inverse length dimensions 5 𝑓1 (𝛿) − 3(1 + 𝛿) 𝑓2t (𝛿)
are [𝐸 MF ] = 𝑑 + 𝑧, [𝑟 1,2 ] = 2, [𝜙] = [𝜑] = (𝑑 + 𝑧 − 2)/2 = cos2 𝛼
5𝜋 2
The integrands 𝜀 ±𝜙, 𝜑 ( 𝑝)
® denote the two lower-branch energy
Í 5 𝑓1 (𝛿) − (1 − 𝛿) 𝑓2e (𝛿)
sin2 𝛼 , (3.6)
eigenvalues of the mean-field Hamiltonian 𝐻MF = 5𝑎=1 (1 + +
5𝜋 2
𝑔1 𝑔2
® 𝑎 + √ 𝜙𝛾45 + √ 𝜑𝛾5 . They are given as
𝑠 𝑎 𝛿)𝑑 𝑎 ( 𝑝)𝛾 ∫ ∞ ∫ "
𝑁 𝑁 dΩ 1 1 𝑐+ + 𝑐 − 2
𝐼5/2 (𝛼, 𝛿) B d𝑥 2𝑥 4 + 𝑥
 (2𝜋) 3 𝑋˜ 3/4 2 𝑋˜ 1/2
∑︁  2  2 0
 3
 2  2
𝜀 ±𝜙, 𝜑 ( 𝑝)
®  2 4
= −  (1 − 𝛿) 𝑝 + 4𝛿 2
𝑑 𝑎 ( 𝑝)
® + √
𝑔1 𝜙
+ √
𝑔2 𝜑
1 𝑐+ 1 𝑐−
 𝑁 𝑁 − − +1
 𝑎=1 8 𝑋˜ 1/2 8 𝑋˜ 1/2
 4

Figure S2(b) shows 𝐼5/2 as function of 𝛼 for a representative

fixed value of 𝛿 = −1/2. The rescaled coefficient hosts three
global minima, lying at angles 𝛼 = 0, 𝜋 (AIAO sector) as well
as 𝛼 = 𝜋/2 (nematic sector). This implies that a coexistence
phase (corresponding to 𝛼 ∉ {0, 𝜋/2, 𝜋, 3𝜋/2}) is not realized
for all values of 𝑟 1′ and 𝑟 2′ . As a consequence, the mean-field
theory predicts a direct discontinuous transition between the
AIAO and nematic orders, in agreement with the numerical
data shown in Fig. 2(a) of the main text.

FIG. S2: (a) Coefficient 𝐼5/2 of the nonanalytic term ∝ 𝜒5/2 in

the mean-field energy, for fixed anisotropy parameter 𝛿 = −1/2, as IV. DERIVATION OF FLOW EQUATIONS
function of polar angle 𝛼 in (𝜙, 𝜑) field space. Here, 𝛼 ∈ {0, 𝜋}
(𝛼 ∈ {𝜋, 3𝜋/2}) corresponds to the AIAO (nematic) sector. 𝐼5/2 is In this supplemental section, we provide details on the
positive for all 𝛼 ∈ [0, 2𝜋), justifying the truncation of the mean-field
derivation of the flow equations.
energy after the nonanalytic term. (b) Rescaled coefficient 𝐼5/2 as
function of rescaled polar angle 𝛼, for fixed 𝛿 = −1/2, unveiling three
global minima at 𝛼 = 0, 𝜋 (AIAO sector) and 𝛼 = 𝜋2 (nematic sector).
This excludes the possibility of coexisting AIAO and nematic order.
A. 4 − 𝝐 expansion

√︂ √︂ #
2 4
𝑐+ 2 2 4
𝑐− 2 To begin with, note that the fermion-boson model defined
− 𝑥 𝑥 + 1/2 𝑥 + 1 − 𝑥 𝑥 + 1/2 𝑥 + 1 . in Eq. (1.4) features a unique upper critical dimension. This
𝑋˜ 𝑋˜
is because the effective charge 𝑒, the AIAO Yukawa coupling
(3.7)
𝑔1 , the nematic Yukawa coupling 𝑔2 , and the nematic self-
The integral 𝐼2 (𝛼, 𝛿) can be computed analytically and leads to interaction 𝜆 become simultaneously marginal in 𝑑 = 4 spatial
anisotropy-dependent renormalizations of the tuning parame- dimensions. The presence of a unique upper critical dimen-
ters 𝑟 1 and 𝑟 2 . The involved dimensionless functions 𝑓1 , 𝑓2e , sion enables a controlled 4 − 𝜖 expansion. To generalize our
and 𝑓2t stem from solid-angle integration and are defined in theory to noninteger spatial dimensions 2 < 𝑑 < 4, we main-
Sec. II. Notably, the coefficient of the quadratic term in 𝜒 in tain the general dimensional scaling of the couplings, but per-
Eq. (3.4) takes the form 𝑟 1′ cos2 𝛼+𝑟 2′ sin2 𝛼 with dimensionless form angular integrations directly in the physical dimension
renormalized tuning parameters 𝑟 1,2 ′ . Consequently, the coef- 𝑑 = 3 [11]. This approach is considered the most suitable
ficient is minimized for 𝛼 = 0, 𝜋 if 𝑟 1′ < 𝑟 2′ and for 𝛼 = 𝜋, 3𝜋/2 for models of this type [7]. By integrating out modes with
momenta 𝑞 in the thin shell Λ/𝑏 < 𝑞 < Λ, where Λ is the
if 𝑟 1′ > 𝑟 2′ . The nonanalytic term 𝐼5/2 (𝛼, 𝛿) 𝜒5/2 arises from the
ultraviolet cutoff, and arbitrary frequencies, the renormalized
presence of gapless fermionic degrees of freedom. For gen-
action takes the form
eral (𝛼, 𝛿), it has to be calculated numerically. Figure S2(a)
shows 𝐼5/2 as function of 𝛼 for a representative fixed value of
∫∞ ∫Λ/𝑏 ( 𝑁 
𝛿 = −1/2. Importantly, 𝐼5/2 (𝛼, 𝛿 = −1/2) > 0 for all values d𝜔 d𝑑 𝑞 ∑︁ † 𝜂1
of 𝛼. This implies a continuous transition between the disor-
<
𝑆FB = 𝜓 𝑏 i𝜔
2𝜋 (2𝜋) 𝑑 𝑖=1 𝑖
dered Luttinger semimetal and the long-range-ordered phases, −∞ 0
such that the higher-order terms O ( 𝜒3 ) in Eq. (3.4) can be ∑︁
5 ∑︁
5 
neglected near the transition. In order to determine the nature +𝑏 𝜂𝜓
® 𝑎 + (𝛿 + Δ𝛿)
𝑑 𝑎 ( 𝑞)𝛾 𝑠 𝑎 𝑑 𝑎 ( 𝑞)𝛾
® 𝑎 𝜓𝑖
of the transition between the two ordered phases, it is use- 𝑎=1 𝑎=1
ful to rescale the fields 𝜙 and 𝜑 in a way that the coefficient 1 1  
of the nonanalytic term in the AIAO sector is the same as + 𝑎𝑏 𝜂𝑎 𝑞 2 𝑎 + 𝜙 𝑏 𝜂 𝜙 𝑞 2 + (𝑟 1 + Δ𝑟 1 ) 𝜙
2 2
)
1 ∑︁
those in the nematic sector. This can be achieved by intro- 5
 𝜂 2 
ducing√the rescaled polar angle 𝛼 using √ the field redefinitions + 2
𝜑 𝑎 𝑏 𝑞 + (𝑐 + Δ𝑐)𝜔 + (𝑟 2 + Δ𝑟 2 ) 𝜑 𝑎
𝜑

𝜙 = [ 𝑁/(𝑎 1 𝑔1 )] 𝜒 cos 𝛼 and 𝜑 = [ 𝑁/(𝑎 2 𝑔2 )] 𝜒 sin 𝛼 with 2 𝑎=1

𝑎 1 B [𝐼5/2 (𝛼 = 0)] 2/5 and 𝑎 2 B [𝐼5/2 (𝛼 = 𝜋/2)] 2/5 . Note ∫∞ ∫Λ/𝑏 𝑑 "
d 𝑞 1 d𝑑 𝑞 2 𝑒 + Δ𝑒 ∑︁ †
𝑁
that 𝛼 = 𝛼 for 𝛼 ∈ {0, 𝜋/2, 𝜋, 3𝜋/2}. On the level of the d𝜔1 d𝜔2
+ i √ 𝑎 𝜓𝑖 𝜓𝑖
quadratic term ∝ 𝜒2 in the mean-field energy, the rescaling (2𝜋) 2 (2𝜋) 2𝑑 𝑁 𝑖=1
can be absorbed in a redefinition of the tuning parameters 𝑟 1 −∞ 0
and 𝑟 2 . The rescaled coefficient of the nonanalytic term ∝ 𝜒5/2 𝑔1 + Δ𝑔1 ∑︁
𝑁
now takes the form + √ 𝜙 𝜓𝑖† 𝛾45 𝜓𝑖
! 5/4
𝑁 𝑖=1
cos2 𝛼 sin2 𝛼 𝑔2 + Δ𝑔2 ∑︁ ∑︁ †
5 𝑁
𝐼5/2 (𝛼, 𝛿) B + 𝐼5/2 (𝛼(𝛼), 𝛿) . (3.8) + √ 𝜑𝑎 𝜓𝑖 𝛾 𝑎 𝜓𝑖
𝑎 21 𝑎 22 𝑁 𝑎=1 𝑖=1
 5

d𝛿 2
= −𝜂 𝜓 𝛿 + (1 − 𝛿2 ) [(1 + 𝛿) 𝑓1t − (1 − 𝛿) 𝑓1e ]𝑒 2
d ln 𝑏 15𝑁
2 𝑔12
− (1 − 𝛿2 ) [(1 + 𝛿) 𝑓1t + (1 − 𝛿) 𝑓1e ]
15𝑁 (1 + 𝑟 1 ) 3
2
2 𝑔2
+ [(1 + 𝛿) 𝑓1t − (1 − 𝛿) 𝑓1e ] , (4.2)
5𝑁 (1 + 𝑟 2 ) 3
 
d𝑒 2 2𝛿 d𝛿
= 𝑧 + 2 − 𝑑 − 𝜂𝑎 + 𝑒2 , (4.3)
d ln 𝑏 1 − 𝛿2 d ln 𝑏
 
d𝑔12 2𝛿 d𝛿
= 6 − 𝑑 − 𝑧 − 𝜂 𝜙 − 2𝜂 𝜓 + 𝑔2
d ln 𝑏 1 − 𝛿2 d ln 𝑏 1
2
+ (1 − 𝛿) (1 − 𝛿2 ) 𝑓2e 𝑔12 𝑒 2
5𝑁
2 𝑔4
− (1 − 𝛿)(1 − 𝛿2 ) 𝑓2e 1
5𝑁 1 + 𝑟1
2 2
2 𝑔 𝑔
− (1 − 𝛿) 𝑓2e 1 2 , (4.4)
5𝑁 1 + 𝑟2
d𝑟 1 4
= (2 − 𝜂 𝜙 )𝑟 1 − (1 − 𝛿) (1 − 𝛿2 ) 𝑓2e 𝑔12 , (4.5)
d ln 𝑏 5
d𝑔22 12 𝑔24
= (6 − 𝑑 − 𝑧 (0) − 𝜂 𝜑(0) − 2𝜂 𝜓(0) )𝑔22 +
d ln 𝑏 5𝑁 1 + 𝑟 2
4 4𝑐 1 𝑔2 𝑔2
+ (1 − 𝛿2 )𝑔22 𝑒 2 − (1 − 𝛿2 ) 1 2 , (4.6)
5𝑁 5𝑁 1 + 𝑟1
d𝜆 1 27 𝜆3
= (6 − 𝑑 − 𝑧 (0) − 3𝜂 𝜑(0) )𝜆 − √
d ln 𝑏 2 4𝑁 𝑐(1 + 𝑟 2 ) 5/2
√
FIG. S3: Feynman diagrams at one-loop order contributing to (a)– 3 3
− 𝑔 , (4.7)
(c) the fermion anomalous dimensions 𝜂1 , 𝜂 𝜓 and the anisotropy pa- 35 2
rameter renormalization Δ𝛿, (d) the Coulomb anomalous dimension √ 2
d𝑐 2 21 𝑐𝜆
𝜂 𝑎 , (e) the AIAO order-parameter anomalous dimension 𝜂 𝜙 and the = (2 − 2𝑧 (0) − 𝜂 𝜑(0) )𝑐 + 𝑔22 + , (4.8)
AIAO tuning parameter renormalization Δ𝑟 1 , (f,g) the nematic order- d ln 𝑏 5 4𝑁 (1 + 𝑟 2 ) 5/2
parameter anomalous dimension 𝜂 𝜑 , the parameter renormalization d𝑟 2 8 21 𝜆2
Δ𝑐, and the nematic tuning parameter renormalization Δ𝑟 2 , (h)–(j) the = (2 − 𝜂 𝜑(0) )𝑟 2 − 𝑔22 − √ , (4.9)
d ln 𝑏 5 𝑁 𝑐(1 + 𝑟 2 ) 3/2
charge renormalizations Δ𝑒, (k)–(m) the AIAO Yukawa vertex renor-
malizations Δ𝑔1 , (n)–(p) the nematic Yukawa vertex renormalizations with the anomalous dimensions given by
Δ𝑔2 , (q)–(s) the self-interaction renormalizations Δ𝜆. The contribu-
tions to the flow of 𝑒 2 from (a) and (h), from (b) and (i), and from (c) 2
𝜂𝜓 = (1 − 𝛿2 ) [(1 + 𝛿) 𝑓1t + (1 − 𝛿) 𝑓1e ]𝑒 2
and (j), respectively, cancel as a consequence of a Ward identity. 15𝑁
2 𝑔12
− (1 − 𝛿2 ) [(1 + 𝛿) 𝑓1t − (1 − 𝛿) 𝑓1e ]
15𝑁 (1 + 𝑟 1 ) 3
# 2 𝑔22
𝜆 + Δ𝜆 ∑︁
5 + [(1 + 𝛿) 𝑓1t + (1 − 𝛿) 𝑓1e ] , (4.10)
+ √ tr(Λ𝑎 Λ𝑏 Λ𝑐 )𝜑 𝑎 𝜑 𝑏 𝜑 𝑐 . (4.1) 5𝑁 (1 + 𝑟 2 ) 3
𝑁 𝑎,𝑏,𝑐=1 𝑔22
4 4
𝜂 𝜓(0) = (1 − 𝛿2 )𝑒 2 + , (4.11)
15𝑁 5𝑁 (1 + 𝑟 2 ) 3
𝜂 𝑎 = 𝑒 2 𝑓𝑒2 , (4.12)
We rescale all frequencies as 𝑏𝑧𝜔 ↦→ 𝜔, with dynamical ex- 𝜂𝜙 = 𝑔12 𝑓𝑔2 , (4.13)
ponent 𝑧 = 2 + 𝜂1 − 𝜂 𝜓 , and all momenta as 𝑏 𝑞® ↦→ 𝑞. ® Further,
we renormalize the fields according to 𝑏 − (2+𝑑+𝑧− 𝜂 𝜓 )/2 𝜓 ↦→ 44 2 21 𝜆2
𝜂 𝜑(0) = 𝑔 + √ , (4.14)
𝜓, 𝑏 − (2+𝑑+𝑧− 𝜂𝑎 )/2 𝑎 ↦→ 𝑎, 𝑏 − (2+𝑑+𝑧− 𝜂 𝜙 )/2 𝜙 ↦→ 𝜙, and 35 2 4𝑁 𝑐(1 + 𝑟 2 ) 5/2
𝑏 − (2+𝑑+𝑧− 𝜂 𝜑 )/2 𝜑 𝑎 ↦→ 𝜑 𝑎 . The effective action then becomes 𝜂1 = 0, (4.15)
of the same form as the original action, but with renormal-
ized parameters 𝛿, 𝑒, 𝑔1 , 𝑟 1 , 𝑔2 , 𝜆, 𝑐, and 𝑟 2 . We take into and the dynamical exponent takes the form
account all one-loop self-energy and vertex diagrams depicted
in Fig. S3. The resulting flow equations read 𝑧 = 2 − 𝜂𝜓, 𝑧 (0) = 2 − 𝜂 𝜓(0) . (4.16)
 6

In the above flow equations and anomalous dimensions, we

have rescaled the couplings as (𝑒 2 , 𝑔12 )Λ− 𝜖 /(2𝜋 2 ) ↦→ (1 −
𝛿2 ) (𝑒 2 , 𝑔12 ), which turned out convenient to assess the proper-
ties of the AIAO fixed point [6]. We have further introduced
the dimensionless couplings (𝑔22 , 𝜆2 )Λ− 𝜖 /(2𝜋 2 ) ↦→ (𝑔22 , 𝜆2 ),
the dimensionless tuning parameters (𝑟 1 , 𝑟 2 )Λ−2 ↦→ (𝑟 1 , 𝑟 2 ),
and the dimensionless parameter 𝑐Λ2 ↦→ 𝑐. The dimensionless
functions 𝑓𝑖 (𝛿) are defined in Sec. II. Note that the self-energy
contributions to the flow of 𝑒 2 , depicted in Figs. S3(a)-(c), can-
cel with the explicit charge vertex renormalizations, shown in
Figs. S3(h)-(j), as a consequence of the Ward identity associ-
ated with the minimalistic gauge transformation 𝜓 ↦→ ei𝑒𝜆( 𝜏 ) 𝜓,
𝑎 ↦→ 𝑎 − 𝜕𝜏 𝜆(𝜏) [2, 6]. The parameter 𝑐 is an irrelevant pa-
rameter that is kept in order to regularize purely bosonic loops
in the nematic sector, shown in Figs. S3(g) and S3(s). Note
that such regularization parameter is not required in the AIAO
sector [6], since a cubic bosonic self-interaction is forbidden
by the Ising symmetry of the AIAO order parameter field 𝜙.
In the above, we have expanded the flow equations assuming
small 𝑐 ≪ 1. As will be shown in Sec. V, this assumption
is consistent with 𝑐 being small at all fixed points of interest,
namely, the LAB fixed point, the nematic fixed point, and the
AIAO fixed point.
Following Ref. [2], we neglect the anomalous contribution
Í
∝ 𝑖,𝑑 𝑗,𝑘=1 𝜕𝑖 𝑇𝑖 𝑗 𝜕𝑘 𝑇𝑘 𝑗 in the nematic sector, allowed whenever
the spatial dimension 𝑑 agrees with dim 𝑇, the dimension of
Í (𝑑−1) (𝑑+2)/2 FIG. S4: Complete set of four-fermion box diagrams at the one-loop
the nematic tensor field 𝑇𝑖 𝑗 = 𝑎=1 𝜑 𝑎 Λ𝑎,𝑖 𝑗 . order. In the dynamical bosonization scheme, diagrams (a)-(f) and
For simplicity, the flow equations for the nematic Yukawa (m)-(o) contribute to the flow of the nematic Yukawa coupling 𝑔2 . In
coupling 𝑔22 , the self-interaction coupling 𝜆, the regularization contrast, diagrams (g)-(i) and (j)-(l) do not contribute to the nematic
parameter 𝑐, and the nematic tuning parameter 𝑟 2 have further channel when adding the individual contributions.
been expanded to lowest order in the anisotropy parameter 𝛿
around the isotropic case 𝛿 = 0. This is sufficient to describe
the nematic sector, see Sec. V. For consistency, before any eight-dimensional subspace spanned by 𝛿, 𝑒, 𝑔1 , 𝑟 1 , 𝑔2 , 𝜆,
rescalings of coupling constants, also the involved anomalous 𝑐, and 𝑟 2 . The cost of this simplification is the introduction
dimensions 𝜂 𝜓 and 𝜂 𝜑 and the dynamical exponent 𝑧 have of an additional free parameter 𝑐 1 in the flow of the nematic
to be expanded to their leading orders in 𝛿 in these cases, Yukawa coupling 𝑔2 . However, the contribution involving 𝑐 1
𝜂 𝜓(0) = 𝜂 𝜓 𝛿=0 , 𝜂 𝜑(0) = 𝜂 𝜑 𝛿=0 , and 𝑧 (0) = 𝑧 𝛿=0 . Note, how- is merely a biquadratic term, ∝ 𝑔12 𝑔22 . As such, we can already
anticipate that its impact on the flow will likely be quantitative
ever, that anisotropy-dependent coupling reparametrizations,
rather than qualitative. Indeed, as will be shown in Sec. V,
such as the above rescaling (𝑒 2 , 𝑔12 ) ↦→ (1 − 𝛿2 ) (𝑒 2 , 𝑔12 ), can
both the location as well as the scaling exponents of the LAB
still lead to crucial anisotropy dependencies. On the level of
fixed point, the nematic fixed point, and the AIAO fixed point
the 4 − 𝜖 expansion, this happens explicitly through two vertex
remain entirely unaffected by the value of 𝑐 1 . We therefore
corrections in the flow equation of 𝑔22 now involving a factor
expect that disregarding the distinction between the T2g and
of 1 − 𝛿2 and implicitly in the flow equations of 𝑔22 , 𝜆, and 𝑐 Eg components of the nematic tensor order parameter will not
through the reparametrized anomalous dimension 𝜂 𝜓(0) involv- affect the universal physics. For the numerical integration of
ing a factor of 1−𝛿2 , see Eq. (4.11). In general, the theory space the flow equations presented in Sec. VI, we choose 𝑐 1 ≡ 1/5,
is spanned by 13 different parameters, namely the anisotropy which represents the weighted average of the T2g and Eg con-
parameter 𝛿, the effective charge 𝑒, the AIAO Yukawa cou- tributions 𝑐 1t = +1 and 𝑐 1e = −1.
pling 𝑔1 , the AIAO tuning parameter 𝑟 1 , two nematic Yukawa The flow equations presented here generalize those previ-
couplings 𝑔2,t and 𝑔2,e (associated with the T2g and Eg com- ously derived for the individual AIAO [6] and nematic [2]
ponents of the nematic tensor order parameter), three cubic sectors, reducing to these in the respective limits.
nematic self-interactions 𝜆ttt , 𝜆tte , and 𝜆eee , two regularization
parameters 𝑐 t and 𝑐 e , and two nematic tuning parameters 𝑟 2,t
and 𝑟 2,e . Note that T2g ⊗ Eg ⊗ Eg = 2T1g ⊕ 2T2g [12], which B. Dynamical bosonization scheme
forbids terms like 𝜑1 𝜑4 𝜑5 in the action. This implies the ab-
sence of a fourth cubic self-interaction governed by a coupling In three spatial dimensions, it is no longer sufficient to solely
𝜆tee . Expanding the flow in the nematic sector around the work with the flow equations arising from the 4 − 𝜖 expansion.
isotropic limit allows us to restrict the parameter space to an Additionally, we need to take into account four-fermion box
 7

diagrams, which generate further four-fermion interactions, form

not present at the microscopic scale. These are shown at the d𝑔22 12 𝑔24
one-loop order in Fig. S4. Although these contributions can = (6 − 𝑑 − 𝑧 (0) − 𝜂 𝜑(0) − 2𝜂 𝜓(0) )𝑔22 +
be safely neglected for small 𝜖 = 4 − 𝑑 ≪ 1, four-fermion d ln 𝑏 5𝑁 1 + 𝑟 2
contributions to the nematic channel turn out to be crucial in 4 4𝑐 1 𝑔2 𝑔2
2 < 𝑑 < 4 [2]. Neglecting these contributions would mask + (1 − 𝛿2 )𝑔22 𝑒 2 − (1 − 𝛿2 ) 1 2
5𝑁 5𝑁 1 + 𝑟1
the fixed-point collision and annihilation of the LAB fixed 𝑟 2 𝑔14
1 1
point with the nematic fixed point when lowering the number + (1 − 𝛿2 ) 2 𝑟 2 𝑒 4 + (1 − 𝛿2 ) 2
of quadratic band touching points 𝑁 below a critical flavor 10𝑁 10𝑁 (1 + 𝑟 1 ) 2
number 𝑁c . Using dynamical bosonization in the nematic 13 𝑟 2 𝑔24 𝑐2 2
𝑟 2 𝑔12 𝑔22
sector, RG-generated four-fermion interactions can be directly + − (1 − 𝛿 ) ,
10𝑁 (1 + 𝑟 2 ) 2 𝑁 (1 + 𝑟 1 )(1 + 𝑟 2 )
incorporated into the flow of the nematic Yukawa coupling (4.19)
𝑔22 [2, 13–16]. To illustrate this, consider the renormalized
action where the first four summands stem from the 4 − 𝜖 expansion
above and the last four terms are the contributions arising
∫ ∫ ∫ from the dynamical bosonization scheme. Restricting to the
1 𝑔2 + Δ𝑔2 eight-dimensional parameter space required us to introduce
<
𝑆FB = 𝜑 𝑎 (𝑟 2 + Δ𝑟 2 )𝜑 𝑎 + √ 𝜑 𝑎 𝜓𝑖† 𝛾 𝑎 𝜓𝑖 another free parameter 𝑐 2 . As with the previously encountered
2 𝑁
parameter 𝑐 1 , the contribution corresponding to 𝑐 2 is also
∫ ∫ ∫
𝜔, 𝑞® 𝜔1 , 𝑞®1 𝜔2 , 𝑞®2
Δ𝐺 2  † 2 simply biquadratic, ∝ 𝑔12 𝑔22 . We show in Sec. V that the
+ 𝜓𝑖 𝛾 𝑎 𝜓𝑖 + . . . , (4.17) location and the scaling exponents of the LAB fixed point, the
𝑁
𝜔1 , 𝑞®1 𝜔2 , 𝑞®2 𝜔3 , 𝑞®3 nematic fixed point, and the AIAO fixed point are independent
of both 𝑐 1 and 𝑐 2 . For the numerical integration of the flow
equations, we choose, in analogy to the procedure for 𝑐 1 , the
∫ ∫ ∞ d𝜔 ∫ Λ/𝑏 d𝑑 𝑞 weighted average 𝑐 2 ≡ 2/5 of the T2g and Eg contributions
where B and the ellipsis denotes all
−∞ 2 𝜋 0 (2 𝜋 ) 𝑑 𝑐 2t = 0 and 𝑐 2e = 1. The flow equations in the dynamical
𝜔, 𝑞®
other, tree level or RG generated, terms. For notational sim- bosonization scheme reduce to the corresponding equations
plicity, we assume summation convention over repeated in- from Ref. [2] in the nematic sector when setting 𝑔1 = 𝛿 = 0.
dices in this section. The renormalizations Δ𝑟 2 , Δ𝑔2 , Δ𝐺 2 =
O (ln 𝑏) contribute to the nematic order-parameter mass, the
V. FIXED-POINT STRUCTURE
nematic Yukawa vertex, and the four-fermion coupling in the
nematic channel, respectively. Note that although the four-
fermion interaction is absent in the ultraviolet, it gets gener- In this supplemental section, we provide details of the dif-
ated under an RG transformation. Shifting the nematic fields ferent RG fixed points. In particular, we show that the LAB
as 𝜑 𝑎 ↦→ 𝜑 𝑎 + √Δ𝑤 𝜓𝑖† 𝛾 𝑎 𝜓𝑖 results in the shifted renormalized fixed point collides (annihilates) with the nematic fixed point
𝑁 at (below) a critical number of quadratic band touching points
action 𝑁c > 1. Moreover, we demonstrate that the AIAO fixed point
does not take part in the collision and annihilation process, but
∫ remains in the physical space of real couplings for any number
1
<
𝑆FB ↦→ 𝜑 𝑎 (𝑟 2 + Δ𝑟 2 )𝜑 𝑎 of quadratic band touching points 𝑁 ∈ N.
2 To observe the fixed-point annihilation at finite coupling
∫ ∫
𝜔, 𝑞®
strengths, it is convenient to reparametrize the AIAO Yukawa
𝑔2 + Δ𝑔2 + 𝑟 2 Δ𝑤
+ √ 𝜑 𝑎 𝜓𝑖† 𝛾 𝑎 𝜓𝑖 coupling 𝑔12 ↦→ 𝑔12 /(−𝛿) and the corresponding AIAO tuning
𝑁 parameter 𝑟 1 ↦→ 𝑟 1 /(−𝛿). Additionally, we introduce the new
∫ ∫ ∫
𝜔1 , 𝑞®1 𝜔2 , 𝑞®2
Δ𝐺 2 + 𝑔2 Δ𝑤  † 2 tuning variables
+ 𝜓𝑖 𝛾 𝑎 𝜓𝑖 + . . . , 𝑟1 𝑟2
𝑁 𝛼1 B , 𝛼2 B , (5.1)
𝜔1 , 𝑞®1 𝜔2 , 𝑞®2 𝜔3 , 𝑞®3 1 + 𝑟1 1 + 𝑟2
(4.18) which map the AIAO and nematic tuning parameters 𝑟 1
and 𝑟 2 for 𝑟 1 , 𝑟 2 ∈ [0, ∞) onto the unit interval, 𝛼1 , 𝛼2 ∈
[0, 1). The fixed-point equations for the eight parameters
assuming Δ𝑤 = O (ln 𝑏). Choosing Δ𝑤 = −Δ𝐺 2 /𝑔2 elimi- 𝛿, 𝑒 2 , 𝑔12 , 𝛼1 , 𝑔22 , 𝜆, 𝑐, 𝛼2 can be solved analytically in the limit
nates the four-fermion interaction at the cost of the additional 𝑁 → ∞. Starting from the large-𝑁 solutions, we use a
Yukawa vertex renormalization ∝ −𝑟 2 Δ𝐺 2 /𝑔2 . In other words, Newton-Raphson method in order to determine the fixed-point
the RG-generated four-fermion interaction can be incorporated values and the corresponding scaling or critical exponents nu-
through a Hubbard-Stratonovich transformation after each RG merically at finite 𝑁. For sufficiently large 𝑁, we identify a
step, resulting in a modified flow equation for the Yukawa unique interacting infrared stable fixed point, which we as-
coupling. Within the dynamical bosonization scheme, the sociate with the disordered LAB phase [17–19]. The corre-
flow equation of the nematic Yukawa coupling 𝑔22 takes the sponding fixed-point values for different values of 𝑁, together
 8

TABLE II: LAB fixed point: Fixed-point values and corresponding scaling exponents in 𝑑 = 3 spatial dimensions from dynamical bosonization
for different numbers of quadratic band touching points 𝑁 ≥ 𝑁c = 1.856. Below the critical number 𝑁c , the LAB fixed point has annihilated
with the nematic fixed point and is no longer present in the physical space of real couplings.

𝑁 𝛿★ 2
𝑒★ 2
𝑔1★ 𝛼1★ 2
𝑔2★ 𝜆★ 𝑐★ 𝛼2★ 2−𝑧 𝜂𝑎 𝜂𝜙 𝜂𝜑 𝜔1 𝜔2
𝑁c 0.000 0.874 0.732 1.000 1.446 −0.040 0.162 0.927 0.126 0.874 1.004 1.818 0.036 0.000
2 0.000 0.882 0.737 1.000 1.494 −0.040 0.164 0.952 0.118 0.882 1.011 1.878 0.034 0.269
3 0.000 0.918 0.738 1.000 1.542 −0.040 0.163 0.976 0.082 0.918 1.012 1.938 0.023 0.624
4 0.000 0.937 0.737 1.000 1.556 −0.040 0.162 0.983 0.063 0.937 1.011 1.956 0.018 0.740
5 0.000 0.949 0.736 1.000 1.564 −0.040 0.162 0.987 0.051 0.949 1.009 1.966 0.014 0.800
10 0.000 0.974 0.733 1.000 1.578 −0.040 0.161 0.994 0.026 0.974 1.005 1.984 0.007 0.908
25 0.000 0.989 0.731 1.000 1.586 −0.040 0.160 0.997 0.011 0.989 1.002 1.994 0.003 0.965
100 0.000 0.997 0.730 1.000 1.590 0.159 0.999 0.003 0.997 1.001 1.998 0.001 0.991
√︃
−0.040
∞ 0 1 35 1 35 105
− 13310 7 1 0 1 1 2 0 1
48 22 44

TABLE III: Nematic fixed point: Fixed-point values and corresponding critical exponents in 𝑑 = 3 spatial dimensions from dynamical
bosonization for different numbers of quadratic band touching points 𝑁 ≥ 𝑁c = 1.856. Below the critical number 𝑁c , the nematic fixed point
has annihilated with the LAB fixed point and is no longer present in the physical space of real couplings.

𝑁 𝛿★ 2
𝑒★ 2
𝑔1★ 𝛼1★ 2
𝑔2★ 𝜆★ 𝑐★ 𝛼2★ 2−𝑧 𝜂𝑎 𝜂𝜙 𝜂𝜑 1/𝜈
𝑁c 0.000 0.874 0.732 1.000 1.446 −0.040 0.162 0.927 0.126 0.874 1.004 1.818 0.000
2 0.000 0.882 0.725 1.000 1.378 −0.039 0.158 0.892 0.118 0.882 0.994 1.732 0.260
3 0.000 0.916 0.717 1.000 1.215 −0.038 0.145 0.805 0.084 0.916 0.983 1.528 0.577
4 0.000 0.934 0.716 1.000 1.125 −0.037 0.137 0.755 0.066 0.934 0.981 1.414 0.681
5 0.000 0.946 0.716 1.000 1.065 −0.037 0.132 0.721 0.054 0.946 0.982 1.339 0.740
10 0.000 0.971 0.720 1.000 0.934 −0.036 0.120 0.644 0.029 0.971 0.988 1.174 0.862
25 0.000 0.988 0.725 1.000 0.851 −0.035 0.112 0.594 0.012 0.988 0.994 1.070 0.943
100 0.000 0.997 0.728 1.000 0.809 0.108 0.569 0.003 0.997 0.999 1.017 0.985
√︃
−0.035
∞ 0 1 35 1 35 105
− 85184 7 14 0 1 1 1 1
48 44 66 25

TABLE IV: AIAO fixed point: Fixed-point value and exact critical exponents in 𝑑 = 3 spatial dimensions. The AIAO fixed point is present in
the physical space of real couplings for any number of quadratic band touching points 𝑁 ∈ N.

2 2 2 2−𝑧 1/𝜈
√︃ ★
𝑁 𝛿★ 𝑒★ 𝑔1★ 𝛼1★ 𝑔2★ 𝜆 𝑐★ 𝛼2★ 𝜂𝑎 𝜂𝜙 𝜂𝜑
all −1 9 9 0 35 − 105 7 1 0 1 1 1 1
16 16 44 85184 66

with the associated scaling exponents, are given in Table II. anomalous dimension as 𝜂 𝜓 = 2 − 𝑧, the photon anomalous
In addition, as expected from the mean-field analysis, we fur- dimension 𝜂 𝑎 , the anomalous dimension 𝜂 𝜙 of the AIAO order
ther find two quantum critical fixed points associated with parameter field, the anomalous dimension 𝜂 𝜑 of the nematic
nematic [2, 7] and AIAO [3, 4, 6] instabilities. Their corre- order parameter, as well as the two least negative eigenvalues
sponding fixed-point values and critical exponents are given of the stability matrix at the LAB fixed point, 𝜔1 and 𝜔2 ,
in Tables III and IV, respectively. corresponding to leading and next-to-leading corrections to
As visible from Tables II and III, the nematic fixed point scaling. (We have labelled 𝜔1 and 𝜔2 such that 𝜔1 < 𝜔2 at
and the LAB fixed point are located in the physical space large 𝑁, and smoothly connected to this limit for finite 𝑁.)
of real couplings for 𝑁 ≥ 𝑁c = 1.8555070 . . . (rounded to The scaling exponent 𝜔1 reflects the fact that the anisotropy 𝛿
three decimal places for better readability in what follows). At becomes an exactly marginal operator in the limit of infinitely
𝑁 = 𝑁c , the LAB fixed point collides with the nematic fixed many quadratic band touching points 𝑁 → ∞. The scaling
point and for even lower numbers of quadratic band touch- exponent 𝜔2 → 0+ for 𝑁 → 𝑁c+ signals the collision of the
ing points 𝑁 < 𝑁c , the two fixed points take complex values LAB fixed point with the nematic fixed point. For the nematic
with finite imaginary parts. This result is consistent with the fixed point, we analogously show 𝑧, 𝜂 𝑎 , 𝜂 𝜙 , and 𝜂 𝜑 , as well
earlier analyses [1, 2, 8]. For the LAB fixed point, we show as the inverse correlation-length exponent 1/𝜈, correspond-
the dynamical exponent 𝑧, which is related to the fermion ing to the unique positive eigenvalue of the stability matrix.
 9

We note that the AIAO order parameter is non-critical at the VI. INTEGRATION OF FLOW EQUATIONS
nematic quantum critical point, and the corresponding anoma-
lous dimension 𝜂 𝜙 at the nematic fixed point determines the In this supplemental section, we give details on the proce-
scaling of the AIAO correlation function only after subtract- dure for obtaining Figs. 2(b–d) and 3 of the main text from
ing the momentum-independent contribution. The numerical numerically integrating out the flow equations, and provide
values of 𝑧, 𝜂 𝑎 , 𝜂 𝜑 , 𝜔2 , and 1/𝜈 in Tables II and III precisely further comments on the universality class associated with the
agree with the corresponding values given in Ref. [2]. This continuous order-to-order transition.
is a consequence of the fact that AIAO fluctuations are sup-
pressed by the infinite mass of the AIAO order-parameter field
𝜙 at the LAB and nematic fixed points, as indicated by the A. Zero-temperature phase diagram
corresponding 𝛼1★ = 1.

As shown in Table IV, the AIAO fixed point exists for any To begin with, we eliminate the Yukawa couplings 𝑔12 and
number of quadratic band touching points 𝑁 ∈ N and, there- 𝑔22by introducing effective couplings 𝐺 1 and 𝐺 2 , correspond-
fore, is not involved in any fixed-point annihilations. The fixed ing to the microscopic couplings in the original four-fermion
point is characterized by maximal anisotropy 𝛿★ = −1, which action [Eq. (2) in the main text], as
has the consequence that the location of the fixed point and
the corresponding critical exponents can be computed analyt- 𝑔12 𝑔22
𝐺1 = and 𝐺2 = , (6.1)
ically. As argued in Sec. VII, the AIAO critical exponents are 𝑟1 𝑟2
expected to be exact. Our results for 𝑧, 𝜂 𝑎 , 𝜂 𝜙 , and 1/𝜈 also
precisely agree with the earlier calculations [4, 6]. cf. Eq. (1.5). To obtain the zero-temperature phase di-
agram, we numerically integrate out the flow equations
We have explicitly verified that locations and exponents of for the eight parameters 𝛿, 𝑒 2 , 𝐺 1 , 𝛼1 , 𝐺 2 , 𝜆, 𝑐, 𝛼2 in 𝑑 =
all three fixed points are independent from the parameters 3 spatial dimensions, using varying initial effective in-
𝑐 1 and 𝑐 2 introduced in Sec. IV. This implies that also our teraction strengths  in the AIAO (nematic) channel
√  𝐺1
result for the critical number 𝑁c at which the fixed-point col- (𝐺 2 ), and fixed 𝛿, 2𝑒 2 /(3𝜋 2 Λ), 𝛼1 , 𝜆/ 2𝜋 2 Λ, 𝑐Λ2 , 𝛼2 =
lision takes place does depend on neither 𝑐 1 nor 𝑐 2 . This  √︃  UV
justifies a posteriori the earlier introduced restriction to an 1
− 12 , 1, 10 105
, − 85184 7 14
, 66 , 25 at the ultraviolet scale. In order
eight-dimensional theory space.
to allow for an easier comparison with the mean-field results
Let us end this section by a brief comparison of the present and other studies, we have chosen to work with dimensionful
fixed-point annihilation in Luttinger semimetals with the fixed- quantities 𝑒 2 , 𝐺 1 , 𝐺 2 , 𝜆, and 𝑐 in this section. The initial values
point annihilation occurring in the 𝑁-component Abelian of the effective charge 𝑒 2 , the cubic nematic self-interaction
Higgs model in three space-time dimensions [20–23]. In both 𝜆, the regularization parameter 𝑐, and the nematic mass pa-
cases, fixed-point annihilations occur by lowering the num- rameter 𝛼2 have been chosen according to the location of the
ber of degrees of freedom below a certain critical number nematic fixed point in the limit of infinitely many quadratic
band 2
𝑁c . However, as established above, Luttinger semimetals re- √ touching points 𝑁 → ∞. The factors of 2/(3𝜋 ) and
alize a collision and annihilation of a critical fixed point with 2
1/ 2𝜋 in the definition of the starting values for 𝑒 and 𝜆 2

a stable fixed point, while the Abelian Higgs model realizes arise from the coupling rescaling for 𝛿 = −1/2, as given after
a collision and annihilation of a critical fixed point with a Eq. (4.16). While the location of the phase boundaries depends
bicritical fixed point. Without additional fine tuning, a con- on the choice of the initial values, our universal results, such
tinuous order-to-order transition in a system described by the as the occurrence of a continuous order-to-order transition is
𝑁-component Abelian Higgs model for 𝑁 < 𝑁c is therefore independent of this choice. Depending on the chosen initial
not possible. This is consistent with numerical results on conditions, for 𝑁 > 𝑁c , we find a flow towards either of three
SU(𝑁) spin models on two-dimensional lattices, which are different regimes: For small 𝐺 1 and 𝐺 2 , the parameters flow
believed to be described by the 𝑁-component Abelian Higgs towards the LAB fixed-point values. In particular, this implies
model [24, 25]. For 𝑁 = 2, the quantum phase transition be- that 𝛼1 and 𝛼2 remain positive at all RG scales. This regime is
tween a Néel antiferromagnet and a valence bond solid appears associated with the LAB phase. For positive 𝐺 1 beyond a cer-
weakly first order, which has been associated with the disap- tain finite threshold, we find that 𝛼1 becomes negative at some
pearance of the quantum critical fixed point in the Abelian RG scale. Negative 𝛼1 corresponds to a negative curvature of
Higgs model due to the fixed-point collision [25–27]. Thus, the effective potential in the AIAO sector. Consequently, we
ironically, it is the very fixed-point collision mechanism that is associate the flow of 𝛼1 toward negative values with the onset
responsible for the continuous order-to-order transition in the of AIAO order. For positive 𝐺 2 beyond a certain finite thresh-
Luttinger semimetals (where it destabilizes the LAB phase) old, on the other hand, we find that 𝛼2 becomes negative at
that leads to a weakly-first-order transition in the SU(𝑁) spin some RG scale, which we associate with the onset of nematic
models (where it destabilizes the required quantum critical order. In order to estimate the values of the order parameters
fixed point). Similar comparisons to fixed-point annihilations in the two different ordered phases, we assume that the flow
in tensor O(𝑁) models [28–30] and 𝑄-state Potts models [31– away from the quantum critical fixed point is dominated by
34] can be made. simple dimensional scaling. This assumption only modifies
 10

d®
𝑟 (𝑡 )
FIG. S5: (a) Flow speed as function of RG time 𝑡 ∈ [0, 𝑡IR ] for 𝑁 = 1 and a set of initial values of the flow parameters on the nematic
d𝑡
√︃
side of the transition, namely 𝑟®(𝑡 = 0) = (− 12 , 1, 3 𝜋 2𝐺/(2Λ) − 34 · 10 −5 , 10
1 , 0, − 105 7 14 3 𝜋2
85184 , 66 , 25 ), with 𝐺 1c = 0.752 . . . · 2Λ . The speed displays
1c

d®
𝑟 (𝑡 )
two regimes of slow flow, indicated by the two minima of d𝑡 . The first (second) regime arises from the vicinity to the AIAO (complex
LAB) fixed point. The local maximum at intermediate 𝑡 LAB corresponds to the crossover between the two regimes. (b) AIAO order parameter
⟨𝜙⟩ as function of 𝐺 1 − 𝐺 1c in double logarithmic plot on the AIAO side of the transition for 𝑁 = 1. All other initial flow parameters have
been chosen as in (a). Dashed line indicates the theoretically expected slope 𝜈𝑧 = 2. (c) Same as (b), but for the nematic order parameter
⟨𝜑⟩ as function of 𝐺 1c − 𝐺 1 on the nematic side of the transition. The exceptionally small value of the nematic order parameter is a result of
the complex LAB fixed point still being reasonable close to the physical space of real couplings at 𝑁 ≲ 𝑁c = 1.856, resulting in a slow flow
and a sizable intermediate-temperature LAB-like regime. (d) Same as (c), but for the crossover temperature 𝑇★LAB between the nonuniversal
high-temperature regime and the intermediate-temperature LAB-like regime on the nematic side of the transition.

nonuniversal properties of the observables; in particular, it B. Finite-temperature phase diagram

does neither change the order of the transition nor the values
of the corresponding critical exponents [16, 35, 36]. Under
this assumption, the values of the order parameters can be In order to obtain the finite-temperature phase diagram for
computed from the RG time 𝑡IR = ln 𝑏 IR , at which the asso- 𝑁 = 1, shown in Fig. 3(b) of the main text, we again assume
ciated parameters 𝛼1 and 𝛼2 , respectively, turn negative. In that the flow away from the quantum critical fixed point is dom-
particular, in the AIAO phase, the order parameter ⟨𝜙⟩ is re- inated by simple dimensional scaling. Up to prefactors of order
lated to 𝑡 IR via ⟨𝜙⟩ ∝ exp −
𝑑+𝑧−2+𝜂 𝜙
𝑡IR = exp(−2𝑡IR ), since one, we can then estimate the critical temperature on the AIAO
2
side of the transition as 𝑇cAIAO ∝ exp(−𝑧𝑡 IR ) = exp(−2𝑡IR ),
𝑧 = 2 and 𝜂 𝜙 = 1 at the AIAO quantum critical fixed point
where 𝑡 IR = ln 𝑏 IR denotes the RG time at which the pa-
for all numbers of quadratic band touching points 𝑁 ∈ N and
rameter 𝛼1 turns negative. Similarly, on the nematic side of
in 𝑑 = 3 spatial dimensions. In the nematic phase, the corre-
the transition, we can estimate the nematic critical tempera-
sponding order parameter ⟨𝜑⟩ can be computed analogously;
ture as 𝑇cnem ∝ exp(−2𝑡 IR ), with 𝑡IR denoting the RG time
however, one has to be careful which dynamical exponent and
at which the parameter 𝛼2 turns negative. For 𝑁 = 1, the
anomalous dimension to incorporate: For 𝑁 = 10 quadratic 
𝑑+𝑧−2+𝜂 𝜑 complex LAB fixed point is still reasonably close to the phys-
band touching points, we identify ⟨𝜑⟩ ∝ exp − 2 𝑡 IR = ical space of real couplings. This results in a regime with
exp(−2.072𝑡IR ), with 𝑧 = 1.971 and 𝜂 𝜑 = 1.174 at the nematic logarithmically slow flow and LAB-like behavior in a sizable
fixed point, see Table III. Likewise, for 𝑁 = 2 quadratic band  finite-temperature regime [1, 2]. The corresponding crossover
touching points, we employ again ⟨𝜑⟩ ∝ exp −
𝑑+𝑧−2+𝜂 𝜑
𝑡 IR = temperature 𝑇★LAB between the nonuniversal high-temperature
2
regime and the LAB-like regime at intermediate temperatures
exp(−2.307𝑡IR ), since 𝑧 = 1.882 and 𝜂 𝜑 = 1.732 at the ne- can be estimated via 𝑇★LAB ∝ exp(−2𝑡LAB ), where 𝑡LAB corre-
matic fixed point, see Table III. However, for the case 𝑁 = 1, sponds to the RG time at which the flow enters the slow-flow
and therewith  after the fixed-point
 annihilation, we identify regime associated with the complex LAB fixed point. A natural
𝑑+𝑧−2+𝜂 𝜙
⟨𝜑⟩ ∝ exp − 2 𝑡 IR = exp(−2𝑡 IR ), with 𝑧 = 2 and way to estimate 𝑡 LAB is to analyze the dimensionless vector of
flow parameters 𝑟®(𝑡) B (𝛿, 2 2 2 2
𝜂 𝜙 = 1 from the annihilation-surviving AIAO fixed point, √ 𝑒 /[2𝜋 (1−𝛿 )Λ], 𝐺 1 Λ/[2𝜋 (1−
which governs the transition into the nematic phase. Using 2 2 2 2
𝛿 )], 𝛼1 , 𝐺 2 Λ/(2𝜋 ), 𝜆/ 2𝜋 Λ, 𝑐Λ , 𝛼2 ) (𝑡) as function of RG
these prescriptions, we arrive at Figs. 2(b–d) and 3(a) in the time 𝑡. Figure S5(a) shows the corresponding flow speed
main text from numerical integration of the flow equations. d®𝑟 (𝑡 )
as function of 𝑡 ∈ [0, 𝑡IR ]. We observe two regimes in
d𝑡
which the flow becomes slow. Comparing the flow parameters
𝑟®(𝑡) with the values of the different fixed points, we identify
the first (second) regime of slow flow to arise from the vicinity
to the AIAO (complex LAB) fixed point. The first minimum in
d®𝑟 (𝑡 )
d𝑡 thus corresponds to the quantum critical regime, while
 11

the second minimum corresponds to the LAB regime. The

local maximum in d®𝑟d𝑡(𝑡 ) in between the two minima at inter-
mediate time 𝑡 LAB thus corresponds to the crossover between
the two regimes.
In sum, on the nematic side of the transition, we find a
sizable intermediate-temperature regime between the critical
temperature 𝑇cnem and the crossover temperature 𝑇★LAB . On FIG. S6: Every loop contribution to the nematic Yukawa vertex falls
the AIAO side, there is only a single minimum in the flow into one of two distinct classes, (a) the class of diagrams which involve
speed d®𝑟d𝑡(𝑡 ) , associated with the quantum critical regime, in- at least one internal nematic boson and (b) the class of diagrams which
do not involve any internal nematic boson. While class (a) does not
dicating the absence of an additional intermediate-temperature contribute at the AIAO fixed point due to infinite nematic boson mass
regime, in agreement with Fig. 3(b) of the main text. 𝑟 2★ = ∞, class (b) vanishes at the AIAO fixed point due to maximal
spatial anisotropy 𝛿★ = −1.

C. Universality class
two different order parameters may in general take on different
As a cross-check of our numerical analysis, we demon- values. The fact that this does not happen in the present case
strate that the observables computed by integrating out the can be understood as a consequence of the maximal anisotropy
flow equations lead to the same critical exponents as those 𝛿★ = −1 at the AIAO fixed point, leading to 𝜂 𝜙 = 𝜂 𝜑 = 1,
obtained from our analytical solutions of the set of fixed-point which, together with the hyperscaling assumption, implies that
equations and the eigenvalue problem of the corresponding all pairs of exponents agree.
stability matrix (Sec. V). Since the continuous order-to-order
transition is governed by the AIAO fixed point, we expect
the correlation-length exponent 𝜈 = 1 and the anomalous di- VII. HIGHER-LOOP CORRECTIONS
mensions of the AIAO and nematic order parameter fields
𝜂 𝜙 = 𝜂 𝜑 = 1. The hyperscaling assumption implies for the In this supplemental section, we demonstrate that the critical
order-parameter exponents 𝛽 𝜙 = 𝛽 𝜑 = 2 on the two sides of exponents at the AIAO fixed point we have found at the one-
the transition. Further, we expect the dynamical critical expo- loop order do not receive higher-loop corrections, and are thus
nent 𝑧 = 2. Figure S5(b) displays the AIAO order parameter as expected to be exact. This is ultimately another consequence
a function of the tuning parameter 𝐺 1 − 𝐺 1c on the AIAO side of the maximal anisotropy 𝛿★ = −1 at the AIAO fixed point.
of the transition. Near the critical point, the values obtained Our argument is independent of the number of quadratic band
from numerical integration (solid line) approach the expected touching points 𝑁 and therefore holds for all 𝑁 ∈ N.
behavior with slope 𝛽 𝜙 = 2 (dashed line), as expected from the
As argued in Ref. [6] for 𝐺 2 = 0, higher-loop corrections
fixed-point analysis. Similarly, Fig. S5(c) displays the nematic
in the AIAO and charge sector are fully suppressed at the
order parameter as a function of 𝐺 1c −𝐺 1 on the nematic side of
AIAO fixed point. This can be understood as a consequence
the transition. While corrections to scaling appear to be more
of the fact that the AIAO fixed point is located at maximal
significant in this case, the leading behavior is still consistent
anisotropy 𝛿★ = −1. Here, it can be seen from the fact that
with the expected slope 𝛽 𝜑 = 2. We attribute the sizable scal-
the fixed point values for 𝑒 2 and 𝑔12 are finite after performing
ing corrections to the slow flow arising from the fixed-point
annihilation and the marginal anisotropy parameter 𝛿, which the anisotropy-dependent reparametrization (𝑒 2 , 𝑔12 ) ↦→ (1 −
leads to quasiuniversal behavior over several length scales [6]. 𝛿2 ) (𝑒 2 , 𝑔12 ), cf. Sec. IV. After this rescaling, every higher-loop
Lastly, Fig. S5(d) depicts the crossover temperature 𝑇★LAB as diagram now comes with a strictly positive power of (1 − 𝛿2 ).
a function of 𝐺 1c − 𝐺 1 on the nematic side of the transition. As a result, all higher-loop corrections are strongly suppressed
Again, the numerical result (solid line) in the vicinity of the in the vicinity of the AIAO fixed point at anisotropy 𝛿★ = −1,
critical point is consistent with the expectation from the fixed- rendering the one-loop flow equations asymptotically exact,
point analysis (dashed line), namely 𝑇★LAB ∝ (𝐺 1c − 𝐺 1 ) 𝜈𝑧 see Ref. [6] for details and Refs. [37, 38] for a different model
with 𝜈𝑧 = 2. with analogous behavior.
Let us end this section with two general comments on the In our extended model for finite 𝐺 1 and 𝐺 2 , the AIAO fixed
universality class of an order-to-order quantum phase transi- point remains to be located at maximal anisotropy 𝛿★ = −1 and
tion caused by the proposed fixed-point annihilation mech- additionally lies now at infinite nematic boson mass 𝑟 2★ → ∞,
anism. First, it is of course no coincidence that the two cf. Table IV. These two fixed-point values suppress all higher-
correlation-length exponents 𝜈 𝜙 and 𝜈 𝜑 associated with the loop diagrams contributing to the flow equations relevant for
AIAO and nematic correlation functions align in our theory, the critical exponents. For instance, all higher-loop corrections
𝜈 = 𝜈 𝜙 = 𝜈 𝜑 . This has to hold for any order-to-order quantum to the flow of the anisotropy 𝛿, to the effective charge 𝑒 2 , and to
phase transition arising from fixed-point annihilation as a con- the AIAO Yukawa coupling 𝑔12 are either strongly suppressed
sequence of the presence of a single divergent length scale at by strictly positive powers of (1 − 𝛿2 ) (these are the already
the annihilation-surviving quantum critical fixed point. Sec- known contributions arising in the theory for 𝐺 1 = 0) and/or by
ond, all other pairs of critical exponents associated with the strictly positive powers of 1/(1 + 𝑟 2 ) (these are new contribu-
 12

tions with at least one internal nematic boson) in the vicinity of annihilation mechanism, which prepares us for the examples
the AIAO fixed point. Consequently, the anomalous dimension discussed below. We note, however, that the limit of infinite
of the fermion 𝜂 𝜓 = 0, the dynamical exponent 𝑧 = 2 − 𝜂 𝜓 = 2, order-parameter masses is in principle insufficient to compute
the anomalous dimension of the photon 𝜂 𝑎 = 4 − 𝑑, and the order parameters in symmetry-broken phases, as required for
anomalous dimension the AIAO boson 𝜂 𝜙 = 4 − 𝑑 are exact explicitly demonstrating the proposed order-to-order transi-
critical exponents of the AIAO fixed point. A similar argument tion. This is because fermionic RG flows diverge in the case
can be made for the new flow equation of the nematic Yukawa of spontaneous symmetry breaking.
coupling 𝑔22 here, since all loop-correction fall again into two We start from the flow equations in the dynamical bosoniza-
distinct classes, as depicted in Fig. S6, which both vanish at the tion scheme in terms of the couplings 𝛿, 𝑒 2 , 𝐺 1 , 𝛼1 , 𝐺 2 , 𝜆,
AIAO fixed point. The first class of diagrams consists of all 𝑐, and 𝛼2 . In the limit of infinite order-parameter masses,
loop-corrections involving at least one internal nematic boson. 𝛼1 , 𝛼2 → 1, the flow equations for 𝛿, 𝑒 2 , 𝐺 1 , and 𝐺 2 become
Since internal nematic bosons give rise to the appearance of independent of 𝜆 and 𝑐. In this limit, we find
strictly positive powers of 1/(1 + 𝑟 2 ), the whole first class of
diagrams is suppressed in the vicinity of the AIAO fixed point d𝛿 2
= −𝜂 𝜓 𝛿 + (1 − 𝛿2 ) [(1 + 𝛿) 𝑓1t − (1 − 𝛿) 𝑓1e ]𝑒 2 ,
located at infinite nematic mass 𝑟 2★ = ∞. Any higher-loop d ln 𝑏 15𝑁
order correction that is not of the above form involves only (8.2)
internal fermion, photon, and/or AIAO boson propagators and 2
 
d𝑒 2𝛿 d𝛿
is, following the argument presented in Ref. [6], expected to be = 𝑧 + 2 − 𝑑 − 𝜂𝑎 + 𝑒2 , (8.3)
suppressed by a strictly positive power of (1 − 𝛿2 ). Therefore, d ln 𝑏 1 − 𝛿 d ln 𝑏
2

also the anomalous dimension of the nematic order parameter d𝐺 1 2

= (𝑧 − 𝑑)𝐺 1 + (1 − 𝛿) (1 − 𝛿2 ) 𝑓2e 𝑒 2 𝐺 1
field 𝜂 𝜑 = 4 − 𝑑 is expected to be exact at the AIAO fixed d ln 𝑏 5𝑁
point. The only exponent left to think about is the correlation- 2(1 − 2𝑁)
length exponent 𝜈. Note that the AIAO and nematic propagator − (1 − 𝛿) 𝑓2e 𝐺 21
5𝑁
display the same exponent 𝜈 = 𝜈 𝜙 = 𝜈 𝜑 , since there is only 2
one divergent length scale present at the AIAO fixed point, − (1 − 𝛿) 𝑓2e 𝐺 1 𝐺 2 , (8.4)
5𝑁
as demonstrated in the previous section. Analogously to the d𝐺 2 37 + 16𝑁 2 4
flow equation of the anisotropy 𝛿, the effective charge 𝑒 2 , and = (𝑧 (0) − 𝑑)𝐺 2 + 𝐺2 + (1 − 𝛿2 )𝑒 2 𝐺 2
d ln 𝑏 10𝑁 5𝑁
the AIAO Yukawa coupling 𝑔12 , higher-loop corrections to the
1 4𝑐 1 + 5𝑐 2
flow of the AIAO mass 𝑟 1 are either strongly suppressed by + (1 − 𝛿2 ) 2 𝑒 4 − 𝐺 1𝐺 2
strictly positive powers of (1 − 𝛿2 ) and/or by strictly positive 10𝑁 5𝑁
powers of 1/(1 + 𝑟 2 ) in the vicinity of the AIAO fixed point. 1
+ 𝐺 2, (8.5)
As a consequence, the stability matrix is of block triangular 10𝑁 1
form with 1/𝜈 = 2 − 𝜂 𝜙 = 𝑑 − 2 as the only (strictly) positive
with the anomalous dimensions
eigenvalue.
2
𝜂𝜓 = (1 − 𝛿2 ) [(1 + 𝛿) 𝑓1t + (1 − 𝛿) 𝑓1e ]𝑒 2 , (8.6)
15𝑁
VIII. FERMIONIC RG FLOW
𝜂 𝑎 = 𝑒 2 𝑓𝑒2 , (8.7)
In this supplemental section, we show that the dynami- the full anisotropy-dependent dynamical exponent 𝑧 = 2 − 𝜂 𝜓 ,
cally bosonized flow equations of our model using Hubbard- and the corresponding form 𝑧 (0) = 2 − 15𝑁 4
(1 − 𝛿2 )𝑒 2 in the
Stratonovich fields 𝜙 and 𝜑, as obtained in Sec. IV, reduce, in isotropic limit 𝛿 → 0, relevant for the flow in the nematic chan-
the limit of infinite order-parameter masses, 𝛼1 , 𝛼2 → 1, to nel 𝐺 2 . To arrive at the above flow equations, we have rescaled
those of a purely fermionic model, obtained without introduc- the AIAO order-parameter mass in the dynamically bosonized
ing order-parameter fields explicitly. The latter is defined by flow equations as 𝑟 1 ↦→ (1 − 𝛿2 )𝑟 1 , such that 𝐺 1 = 𝑔12 /𝑟 1
the action remains finite at the AIAO quantum critical point. The di-
∫ ( !
∑︁5
i𝑒
mensionless couplings occurring in Eqs. (8.2)–(8.7) are then
3
𝑆F = d𝜏 d 𝑥® 𝜓 𝜕𝜏 + †
(1 + 𝑠 𝑎 𝛿)𝑑 𝑎 (−i∇)𝛾 𝑎 + √ 𝑎 𝜓 related to the dimensionful couplings of the microscopic ac-
𝑁 tion, Eq. (8.1), via the rescaling 𝑒 2 ↦→ 2𝜋 2 Λ4−𝑑 (1 − 𝛿2 )𝑒 2 ,
)
𝑎=1
𝐺 1 ↦→ 2𝜋 2 Λ2−𝑑 𝐺 1 , and 𝐺 2 ↦→ 2𝜋 2 Λ2−𝑑 𝐺 2 . For 𝑁 = 1, the
𝐺 2 ∑︁ †
5
1 𝐺1 † flow equations for the anisotropy 𝛿, the charge 𝑒 2 , and the
+ (∇𝑎) 2 − (𝜓 𝛾45 𝜓) 2 − (𝜓 𝛾 𝑎 𝜓) 2 ,
2 2𝑁 2𝑁 𝑎=1 anomalous dimensions 𝜂 𝜓 and 𝜂 𝑎 agree with the previous cal-
(8.1) culation [3]. For all 𝑁 ∈ N, but 𝛿 = 0, the flow equations for
the charge 𝑒 2 and the anomalous dimensions 𝜂 𝜓 and 𝜂 𝑎 agree
and has been studied previously in Refs. [1–3]. The goal of with those of Refs. [1, 2]. The flow of 𝐺 2 also agrees, with the
the comparison is twofold. First, we want to demonstrate that exception of slight differences in the prefactors of the terms
our calculations are consistent with the fermionic ones pre- ∝ 𝐺 1 𝐺 2 and ∝ 𝐺 21 . These can be attributed to the fact that
sented in Refs. [1–3]. Second, we want build some intuition we have performed the dynamical bosonization scheme only in
on fermionic flows with respect to the proposed fixed-point the nematic channel, leading to a Fierz-incomplete calculation.
 13

FIG. S7: (a) RG flow diagram in 𝑑 = 3 spatial dimensions and for 𝑁 = 10 in the point-like limit 𝛼1 , 𝛼2 → 1 [Eqs. (8.2)–(8.7)] in the plane
spanned by 𝐺 1 and 𝐺 2 for fixed (𝛿, 𝑒 2 ) = (0, 15𝑁/(4 + 15𝑁)) (top panel) and (𝛿, 𝑒 2 ) = (−1, 9/16) (bottom panel), respectively. Arrows
indicate flow towards infrared. Besides the infrared stable fixed point associated with the disordered LAB phase (LAB), there are two quantum
critical points associated with instabilities towards nematic (QCPnem ) and AIAO (QCPAIAO ) orders, respectively. (b) Same as (a), but for
𝑁 = 2, illustrating that the LAB and QCPnem fixed point approach each other for decreasing 𝑁. (c) Same as (a), but for 𝑁 = 1.91, illustrating
the collision of the LAB and QCPnem fixed points at a critical 𝑁 > 1. (d) Same as (a), but for 𝑁 = 1, illustrating the runaway flow in the
isotropic plane 𝛿 = 0 (top panel) and the persistence of QCPAIAO in the plane 𝛿 = −1 (bottom panel).

15𝑁
This also explains the (minor) differences in the flow of 𝐺 1 in effective charge 𝑒 2 = 4+15𝑁 . The flow in this plane is depicted
comparison with the corresponding calculation in Refs. [1, 2]. for different values of 𝑁 in Fig. S7 (top row), illustrating
the fixed-point collision. Although all fixed points in this
For 𝑁 > 1.91, the fermionic flow equations feature a fully section are analytically accessible, we refrain from stating their
infrared stable fixed point, associated with the LAB phase, as expressions here, as they are rather lengthy.
well as two quantum critical fixed points, associated with the
instabilities towards nematic and AIAO orders, respectively.
We note that the AIAO fixed point in the point-like limit, in
contrast to our results in the dynamical bosonization scheme
(Sec. V), exhibits an additional marginally relevant direction, Importantly, the AIAO fixed point does not take part in
associated with the flow of the anisotropy parameter 𝛿. This is the collision and annihilation of the LAB fixed point with the
a known artifact of the fermionic RG flow that can be attributed nematic fixed point, but remains in the physical space of real
to the negligence of AIAO order-parameter fluctuations in the couplings for all 𝑁 ∈ N, in agreement with our results in
point-like limit [6]. For 𝑁 = 1.91, the infrared stable LAB the dynamical bosonization scheme. The AIAO fixed point is
fixed point and the nematic quantum critical point collide and located in the plane defined by anisotropy 𝛿 = −1 and effective
9
subsequently disappear into the complex plane for 𝑁 < 1.91, charge 𝑒 2 = 16 . The flow in this plane is depicted for different
in qualitative agreement with the fixed-point structure obtained values of 𝑁 in Fig. S7 (bottom row), illustrating the fact that the
from the dynamically bosonized RG flow, discussed in Sec. V, AIAO fixed point persists for all 𝑁 ∈ N. We conclude, that, at
as well as the earlier works [1–3]. In agreement with the least in the case of Luttinger semimetals, the presented fixed-
dynamically bosonized RG flow, the fixed-point collision takes point annihilation mechanism is also available in the purely
place in the isotropic 𝐺 1 -𝐺 2 plane defined by 𝛿 = 0 and finite fermionic formulation.
 14

IX. FURTHER EXAMPLES OF CONTINUOUS in Eq. (8.1). In order to illustrate the analogy, consider the one-
ORDER-TO-ORDER TRANSITIONS loop flow equations for the QED3 -Gross-Neveu model [41, 47],

In this supplemental section, we provide three further ex- d𝑒 2

= (1 − 𝜂 𝑎 ) 𝑒 2 , (9.2)
amples for continuous order-to-order transitions arising from d ln 𝑏
fixed-point annihilation in interacting many-body systems. d𝐺 1 16 2 8 4 2(2𝑁 − 1) 2
= −𝐺 1 + 𝑒 𝐺 1 + 𝑒2 𝐺 2 − 𝑒4 + 𝐺1
d ln 𝑏 3𝑁 𝑁 𝑁 𝑁
4 6
− 𝐺 22 − 𝐺 1 𝐺 2 , (9.3)
A. Kagome quantum magnets 𝑁 𝑁
d𝐺 2 8 2 2(2𝑁 + 1) 2 2
= −𝐺 2 + 𝑒 𝐺1 − 𝐺 2 − 𝐺 1 𝐺 2 , (9.4)
We start with the 𝐽1 -𝐽2 -𝐽3 Heisenberg spin-1/2 model on d ln 𝑏 3𝑁 3𝑁 𝑁
the kagome lattice, defined in Eq. (3) of the main text, with
antiferromagnetic next-neighbor coupling 𝐽1 > 0 and second- with gauge field anomalous dimension 𝜂 𝑎 = 34 𝑒 2 to leading
and third-nearest-neighbor couplings 𝐽2 /𝐽1 = 𝐽3 /𝐽1 = 𝛼. As order. Above, we employed the rescaling 𝑒 2 ↦→ 2𝜋 2 Λ𝑒 2 ,
argued in the main text, the model features a transition at fixed 𝐺 1 ↦→ 2𝜋 2 Λ−1 𝐺 1 , and 𝐺 2 ↦→ 2𝜋 2 Λ−1 𝐺 2 , where Λ denotes
finite temperature between a Dirac spin liquid for 𝛼 < 0.4 and a the ultraviolet cutoff. Figure S8(a) illustrated the fixed-point
time-reversal-symmetry-broken chiral spin liquid for 𝛼 > 0.4, structure of the model. For sufficiently large 𝑁, the infrared
described by the QED3 -Gross-Neveu model [39–45]. The attractive plane 𝑒 2 = 𝑒★2 = 3/4 hosts a fully infrared stable
QED3 -Gross-Neveu model describes fermionic excitations 𝜓 fixed point (cQED3 ), associated with the conformal phase of
and 𝜓 † arising from fractionalization of the spin degrees of QED3 [47–50]. In the context of the extended Heisenberg spin-
freedom, interacting via an emergent U(1) gauge field 𝑎 𝜇 and 1/2 model on the kagome lattice, the latter can be understood
additional local four-fermion interactions, as a U(1) Dirac spin liquid [39, 51, 52]. In addition, the plane
features two quantum critical fixed points, associated with
∫ "  
i𝑒 1 2 instabilities towards Ising Z2 time reversal symmetry breaking
𝑆QED3 -GN = d 𝑥 𝜓¯ 𝑖 𝛾 𝜇 𝜕𝜇 − √ 𝑎 𝜇 𝜓𝑖 + 𝑓 𝜇𝜈
3
(QCPIsing ) and SU(2𝑁) flavor symmetry breaking (QCPflavor )
𝑁 4 ground states, respectively [48, 50]. On the kagome lattice, the
#
𝐺1 ¯
2 𝐺2
2 former can be understood as chiral spin liquid [39, 40], while
− 𝜓𝑖 𝛾45 𝜓𝑖 − ¯
𝜓𝑖 𝛾 𝜇 𝜓𝑖 . (9.1) the latter is expected to realize valence bond solid order [53].
2𝑁 2𝑁
For decreasing 𝑁, the infrared stable cQED3 fixed point and
the quantum critical QCPflavor fixed point approach each other
Here, 𝜓𝑖 and 𝜓¯ 𝑖 B 𝜓 † 𝛾3 denote 𝑁 = 2 flavors of four- and eventually collide at a critical flavor number 𝑁c . This is
component Dirac spinors in (2 + 1)-dimensional Euclidean illustrated in Figs. S8(b)–(d), which shows the RG flow in the
space-time, 𝑓 𝜇𝜈 = 𝜕𝜇 𝑎 𝜈 − 𝜕𝜈 𝑎 𝜇 is the U(1) field-strength ten- 𝐺 1 -𝐺 2 plane for fixed 𝑒 2 = 𝑒★2 and different values of 𝑁 above,
sor associated with the U(1) gauge field 𝑎 𝜇 , the 4 × 4 matrices at, and below the critical flavor number 𝑁c . For 𝑁 < 𝑁c , the
𝛾 𝜇 satisfy the Clifford algebra, and 𝛾45 B i𝛾4 𝛾5 is the Hermi- cQED3 and QCPflavor fixed points are located in the complex
tian product of the two chiral gamma matrices in the reducible plane, leaving behind a runaway flow in the space of real cou-
four-component spinor representation. For notational simplic- plings. Importantly, the fixed point QCPIsing , associated with
ity, we have assumed the summation convention over repeated the time-reversal-symmetry-breaking instability, remains un-
flavor indices 𝑖 = 1, . . . , 𝑁 and space-time indices 𝜇 = 1, 2, 3. affected by the collision and annihilation process. For 𝑁 < 𝑁c ,
Among others, the model features Z2 time reversal and SU(2𝑁) it distinguishes between two different regimes characterized by
flavor symmetry. runaway flows: Consider the line of starting values given by
In 2+1 dimensions, the gauge coupling 𝑒 2 has positive mass 𝑒 2 = 𝑒★2 and 𝐺 2 = 0 at the ultraviolet scale. For 𝐺 1 > 𝐺 1★,
dimension and flows to a finite value in the infrared. The the flow diverges towards (𝐺 1 , 𝐺 2 ) → (∞, 0) at a finite RG
four-fermion couplings 𝐺 1 and 𝐺 2 are power-counting irrele- scale, indicating spontaneous time reversal symmetry break-
vant, but maybe generated by the finite gauge coupling. On ing. For 𝐺 1 < 𝐺 1★, on the other hand, the flow diverges
the level of mean-field theory, the coupling 𝐺 1 , if tuned be- towards (𝐺 1 , 𝐺 2 ) → (−∞, −∞), which has previously been
yond a certain finite threshold, induces a transition between the interpreted as an instability towards chiral symmetry break-
Dirac spin liquid for 𝐺 1 < 𝐺 1c and the chiral spin liquid for ing [41, 46, 50]. This suggests that QCPIsing governs for
𝐺 1 < 𝐺 1c . The latter is characterized by a finite vacuum ex- 𝑁 < 𝑁c a continuous quantum phase transition on the between
pectation value of the fermion bilinear ⟨𝜓𝑖 𝛾45 𝜓𝑖 ⟩ ≠ 0, which two long-range ordered states that break different symmetries.
breaks time reversal symmetry [46]. The four-fermion cou- At the one-loop order, we find the critical flavor number as
pling 𝐺 1 can thus be understood to parameterize the second- 𝑁c = 6.24. Higher-loop contributions may reduce the num-
and third-nearest-neighbor interactions 𝐽2 = 𝐽3 in the micro- ber, but most analytical estimates [48–50, 54, 55] appear to
scopic model. The coupling 𝐺 2 is generated by the RG flow suggest that the fixed-point collision and annihilation occurs
and is therefore included from the outset. above the physical number 𝑁 = 2 relevant for the kagome
Below, we demonstrate that the flow in the space spanned quantum magnets. (We emphasize, however, that we are not
by 𝑒 2 , 𝐺 1 , and 𝐺 2 is qualitatively analogous to the flow of the aware of clear numerical evidence for the scenario in an un-
corresponding 𝑒 2 , 𝐺 1 , and 𝐺 2 in the Luttinger fermion model biased simulation of the QED3 -Gross-Neveu model, such as
 15

FIG. S8: (a) Schematic fixed-point structure of the QED3 -Gross-Neveu model in the space spanned by the couplings 𝑒 2 , 𝐺 1 , and 𝐺 2 . Arrows
indicate flow towards infrared. For sufficiently large 𝑁, the infrared attractive plane 𝑒 2 = 𝑒★ 2 hosts a fully infrared stable fixed point (cQED ),
3
associated with the conformal phase of QED3 , representing the U(1) Dirac spin liquid state in the kagome Heisenberg spin-1/2 model.
In addition, the plane features two quantum critical fixed points, associated with instabilities towards Z2 time reversal symmetry breaking
(QCPIsing ) and SU(2𝑁) flavor symmetry breaking (QCPflavor ) ground states, respectively. On the kagome lattice, the former can be understood
as chiral spin liquid, while the latter is expected to realize valence bond solid order. The critical manifold with respect to time reversal symmetry
breaking (flavor symmetry breaking) is indicated in blue (green). For decreasing 𝑁, cQED3 and QCPflavor approach each other and eventually
collide at a critical flavor number 𝑁c . (b) RG flow diagram for 𝑁 = 7 in the plane 𝑒 2 = 𝑒★ 2 spanned by the four-fermion couplings 𝐺 and 𝐺 ,
1 2
indicating the locations of the infrared stable cQED3 fixed point and the quantum critical QCPIsing and QCPflavor fixed points. (c) Same as (b),
but for the critical flavor number 𝑁 = 𝑁c = 6.24, indicating the fixed-point collision. (d) Same as (b), but for 𝑁 = 2, relevant for the extended
Heisenberg spin-1/2 model on the kagome lattice. The absence of the cQED3 fixed point implies that the U(1) Dirac spin liquid is unstable
and the QCPIsing fixed point governs a direct continuous transition between valence bond solid order and a time-reversal-symmetry-breaking
chiral spin liquid.

Monte Carlo [56, 57].) If indeed 𝑁c > 2 in the QED3 -Gross- tuning parameter, and we consider the case of small positive
Neveu model, it implies for the extended Heisenberg spin-1/2 𝜖 ∈ (0, 1/2). The above spin-boson model may be realized
model on the kagome lattice that (1) the U(1) Dirac spin liq- as an effective description of a quantum impurity in a two-
uid becomes unstable at the lowest temperatures towards a dimensional antiferromagnet in the vicinity of a quantum crit-
valence bond solid ordered ground state and (2) the quantum ical point [59–61].
phase transition between the valence bond solid, stabilized for The RG flow equations for the two couplings 𝑔 𝑥 𝑦 and 𝑔 𝑧 up
𝛼 < 0.4, and the chiral spin liquid, stabilized for 𝛼 > 0.4, is to two-loop order read [62, 63]
continuous and governed by the QCPIsing fixed point, as illus-
trated in the schematic phase diagram shown in Fig. 4(b) of d𝑔 2𝑥 𝑦
= 𝜖𝑔 2𝑥 𝑦 − 𝑔 2𝑥 𝑦 (𝑔 2𝑥 𝑦 + 𝑔 2𝑧 ) + 𝑔 4𝑥 𝑦 (𝑔 2𝑥 𝑦 + 𝑔 2𝑧 ), (9.7)
the main text. d ln 𝑏
d𝑔 2𝑧
= 𝜖𝑔 2𝑧 − 2𝑔 2𝑥 𝑦 𝑔 2𝑧 + 2𝑔 2𝑥 𝑦 𝑔 4𝑧 , (9.8)
d ln 𝑏
B. Anisotropic spin-boson model
where we have rescaled 𝑔 2𝑥 𝑦 ↦→ Λ 𝜖 𝑔 2𝑥 𝑦 and 𝑔 2𝑧 ↦→ Λ 𝜖 𝑔 2𝑧 . The
As an example for a model featuring a continuous order-to- flow is illustrated for different values of 0 < 𝜖 < 1/2 in Fig. S9.
order transition from fixed-point annihilation that is directly For 𝜖 > 0, both couplings 𝑔 2𝑥 𝑦 and 𝑔 2𝑧 are infrared relevant.
amenable to large-scale numerical simulations, consider the For 𝑔 2𝑥 𝑦 < 𝑔 2𝑧 at the ultraviolet scale, the parameters flow to-
anisotropic spin-boson model, which describes a single spin wards (𝑔 2𝑥 𝑦 , 𝑔 2𝑧 ) → (0, +∞). This regime is associated with an
𝑆® in a fluctuating magnetic field ℎ® with an XXZ anisotropy, Ising ordered phase, in which the spin forms local order with
defined by the Hamiltonian [58] a finite moment along the 𝑧 direction. The 𝑧-localized phase
spontaneously breaks Ising Z2 symmetry [64]. For 𝑔 2𝑥 𝑦 > 𝑔 2𝑧
® (9.5)
Hspin-boson = 𝑔 𝑥 𝑦 (ℎ 𝑥 𝑆 𝑥 + ℎ 𝑦 𝑆 𝑦 ) + 𝑔 𝑧 ℎ 𝑧 𝑆 𝑧 + Hbulk ( ℎ). at the ultraviolet scale, the phase structure depends on the
value of the external tuning parameter 𝜖. This is due to a
Here, 𝑔 𝑥 𝑦 and 𝑔 𝑧 correspond to in-plane and out-of-plane com- fixed-point annihilation that takes place at a critical value of
ponents of the 𝑔 tensor, respectively, and the bulk Hamiltonian the tuning parameter 𝜖c = 1/4. For 𝜖 < 𝜖c = 1/4 and 𝑔 2𝑥 𝑦
® is chosen such that the fluctuations of the magnetic
Hbulk ( ℎ) above, but not too far from 𝑔 2𝑧 , the flow is attracted by a fully
field are Gaussian, with power-law correlation infrared stable fixed point, corresponding to an interacting dis-
𝛿 𝑎𝑏 ordered phase (D). For large 𝑔 2𝑥 𝑦 , the parameters flow towards
⟨T𝜏 ℎ 𝑎 (𝜏)ℎ 𝑏 (0)⟩ ∝ , (9.6) (𝑔 2𝑥 𝑦 , 𝑔 2𝑧 ) → (+∞, 0). This regime is associated with an XY
|𝜏| 2− 𝜖
ordered phase, in which the spin forms local order with a fi-
where T𝜏 corresponds to the time ordering operator in imag- nite moment perpendicular to the 𝑧 direction. For 𝜖 = 𝜖 c , the
inary time 𝜏. The bath exponent 𝜖 represents an external fixed point D collides with the quantum critical fixed point
 16

FIG. S9: (a) RG flow diagram of the anisotropic spin-boson model for 𝜖 = 0.2 in the plane spanned by 𝑔 2𝑥 𝑦 and 𝑔 2𝑧 . Arrows indicate flow
towards infrared. For 𝑔 2𝑥 𝑦 < 𝑔 2𝑧 at the ultraviolet scale, the parameters flow towards (𝑔 2𝑥 𝑦 , 𝑔 2𝑧 ) → (0, +∞), associated with an Ising ordered
phase. For 𝑔 2𝑥 𝑦 above, but not too far from 𝑔 2𝑧 , the flow is attracted by the fully infrared stable fixed point D, corresponding to an interacting
disordered phase. For large 𝑔 2𝑥 𝑦 , the parameters flow towards (𝑔 2𝑥 𝑦 , 𝑔 2𝑧 ) → (+∞, 0), associated with an XY ordered phase. (b) Same as (a),
but for 𝜖 = 𝜖 c = 1/4, illustrating the collision of the quantum critical fixed point QCPXY and the infrared stable fixed point D. (c) Same as (a),
but for 𝜖 = 0.3. For all values of 𝑔 2𝑥 𝑦 > 𝑔 2𝑧 , the parameters flow towards (𝑔 2𝑥 𝑦 , 𝑔 2𝑧 ) → (+∞, 0), associated with the 𝑥𝑦-localized phase. The
quantum critical fixed point QCPIsing on the isotropic axis 𝑔 2𝑥 𝑦 = 𝑔 2𝑧 governs a continuous transition between the two different orders.

associated with the onset of XY order. Now for 𝜖 > 𝜖 c , the sions (QCD4 ), supplemented with local four-fermion interac-
parameters flow towards (𝑔 2𝑥 𝑦 , 𝑔 2𝑧 ) → (+∞, 0) for all values of tions, defined by the action [68–73]
𝑔 2𝑥 𝑦 > 𝑔 2𝑧 . Most importantly, on the isotropic axis 𝑔 2𝑥 𝑦 = 𝑔 2𝑧 , ∫ " #
1 1 ∑︁
4
there is a quantum critical fixed point QCPIsing that governs a 𝑆QCD4 -GN = d 𝑥 i𝜓¯ 𝐷𝜓
4 𝜇𝜈
/ + 𝐹𝑧 𝐹𝜇𝜈 + 𝑧
𝐺 𝛼 O𝛼
continuous transition between Ising order and the disordered 4 2 𝛼=1
phase for 0 < 𝜖 < 𝜖c and between the Ising order and XY (9.9)
order phase for 𝜖c < 𝜖 < 1/2. For 𝜖 above, but not too far
from 𝜖 c , the two-loop flow equations thus suggest a continuous with covariant derivative 𝐷 / B 𝛾 𝜇 (𝜕𝜇 − i𝑔 𝐴 𝜇 ), SU(𝑁color )
order-to-order transition governed by the QCPIsing fixed point. gauge field 𝐴 𝜇 B 𝐴 𝑧𝜇 𝑇 𝑧 , and the associated field-strength
In fact, the qualitative picture indicated by the perturbative cal- tensor 𝐹𝜇𝜈 B 𝐹𝜇𝜈 𝑧 𝑇 𝑧 , where 𝑇 𝑧 denote the generators of
culation has recently been confirmed in large-scale numerical SU(𝑁color ). A Fierz-complete basis of local four-fermion in-
calculations [64, 65]; see also Refs. [66, 67] for recent analyt- teractions is given by [68]
ical approaches based on large-𝑆 expansions. The numerical  2  2
calculations show that the true critical value of the control pa- O1 B 𝜓¯ 𝑎 𝛾 𝜇 𝜓 𝑏 + 𝜓¯ 𝑎 𝛾 𝜇 𝛾5 𝜓 𝑏 , (9.10)
rameter 𝜖, at which the quantum critical fixed point QCPXY  2  2
and the infrared stable fixed point D collide, is 𝜖c = 0.2294(1), O2 B 𝜓¯ 𝑎 𝜓 𝑏 − 𝜓¯ 𝑎 𝛾5 𝜓 𝑏 , (9.11)
and thus close to our perturbative estimate. Interestingly, the

2
2
numerical data for the transition between the Ising and XY O3,4 B 𝜓𝛾 ¯ 𝜇 𝜓 ± 𝜓𝛾 ¯ 𝜇 𝛾5 𝜓 . (9.12)
ordered phases (see Fig. 9(d) of Ref. [64]) confirm our the-
oretical predictions for a continuous order-to-order transition For notational simplicity, we have abbreviated 𝜓𝜓 ¯ ≡ 𝜓¯ 𝑎 𝜓 𝑎 and
𝑖 𝑖
from fixed-point annihilation, as formulated in the main text: ( 𝜓¯ 𝜓 ) ≡ 𝜓¯ 𝑖 𝜓𝑖 𝜓¯ 𝑗 𝜓 𝑗 , and assumed summation convention
𝑎 𝑏 2 𝑎 𝑏 𝑏 𝑎

First, on the XY side of the transition, the corresponding or- over repeated color indices 𝑖, 𝑗 and flavor indices 𝑎, 𝑏. Con-
der parameter is significantly suppressed in comparison with sider the so-called Veneziano limit 𝑁flavor , 𝑁color → ∞ with
its counterpart on the Ising side. Second, finite-temperature fixed ratio 𝑥 B 𝑁flavor /𝑁color . In this limit, the flow equations
effects are significantly more enhanced on the XY side of the reduce to a set of three equations [71–73],
transition. Both properties can be understood as direct con-
sequence of the annihilation between the QCPXY and D fixed d𝑔 2 2(11 − 2𝑥) 4 2(34 − 13𝑥) 6
= 𝑔 + 𝑔
points and the resulting slow flow on the XY side of the tran- d ln 𝑏 3 3
sition. − 2𝑥𝑔 4 𝐺 1 , (9.13)
d𝐺 1 𝑥 3
= −2𝐺 1 − (1 + 𝑥)𝐺 21 − 𝐺 22 + 𝑔 4 , (9.14)
C. Quantum chromodynamics with four-fermion interactions
d ln 𝑏 4 4
d𝐺 2 9
= −2𝐺 2 + 2𝐺 22 − 2𝑥𝐺 1 𝐺 2 + 6𝑔 2 𝐺 2 + 𝑔 4 , (9.15)
d ln 𝑏 2
As a final example for a continuous order-to-order tran-
sition, we discuss a model relevant to high-energy physics, where we have rescaled the gauge coupling 𝑔 2 𝑁color /(16𝜋 2 ) →
↦
namely quantum chromodynamics in 3 + 1 space-time dimen- 𝑔 2 and the four-fermion couplings 𝐺 1,2 Λ2 𝑁color /(4𝜋 2 ) → ↦
 17

FIG. S10: (a) Schematic fixed-point structure of QCD4 with additional four-fermion interactions in the Veneziano limit 𝑁flavor , 𝑁color → ∞,
with fixed ratio 𝑥 = 𝑁flavor /𝑁color < 11/2, in the space spanned by the couplings 𝑔 2 , 𝐺 1 , and 𝐺 2 . Arrows indicate flow towards infrared. For
𝑥 below, but not too far from 11/2, the infrared attractive plane 𝑔 2 = 𝑔★ 2 (𝐺 ) = (11 − 2𝑥 − 3𝑥𝐺 )/(13𝑥 − 34) hosts a fully infrared stable
1 1
Caswell-Banks-Zaks fixed point (CBZ), as well as two quantum critical fixed points (QCP1 and QCP2 ). The corresponding critical manifolds
are indicated in blue and green, respectively. For decreasing 𝑥, CBZ and QCP2 approach each other and eventually collide at a critical 𝑥c .
(b) RG flow diagram for 𝑥 = 4.3 in the plane 𝑔 2 = 𝑔★2 spanned by the four-fermion couplings 𝐺 and 𝐺 , indicating the locations of the infrared
1 2
stable CBZ fixed point and the quantum critical fixed points QCP1 and QCP2 . (c) Same as (b), but for the critical ratio 𝑥 = 𝑥c = 4.05, indicating
the fixed-point collision. (d) Same as (b), but for 𝑥 = 3.7. The absence of the stable CBZ fixed point implies a direct continuous transition
between a chiral-symmetry-broken state, signaled by the runaway flow towards 𝐺 2 → +∞, and a different ordered ground state, signaled by
the runaway flow towards (𝐺 1 , 𝐺 2 ) → (−∞, −∞).

𝐺 1,2 . For reasons of consistency with the previous sections, points (QCP1 and QCP2 ). For decreasing 𝑥, the infrared stable
we write the flow equations in terms of an RG scale ln 𝑏, with CBZ fixed point and one of the two quantum critical points
ln 𝑏 = 0 (ln 𝑏 → ∞) corresponding to the ultraviolet (infrared) (QCP2 ) approach each other and eventually collide at the crit-
scale. ical value 𝑥c = 4.05. This is illustrated in Figs. S10(b)–(d),
The fixed-point structure of the model in the Veneziano which show the RG flow in the plane 𝑔 2 = 𝑔★2 (𝐺 1 ) for different
limit is shown schematically in Fig. S10(a). For 𝑥 > 11/2, values of 𝑥. The runaway flow towards 𝐺 2 → +∞ at finite
asymptotic freedom is lost and the gauge coupling vanishes RG scale is usually associated with an instability towards a
in the infrared if 𝐺 1 and 𝐺 2 are small at the ultraviolet chiral symmetry breaking ground state [69, 70, 76]. The other
scale. For 𝑥 < 11/2, the gauge charge is a RG relevant quantum critical point (QCP1 ) in the infrared attractive plane
parameter and flows towards a finite value 𝑔 2 = 𝑔★2 (𝐺 1 ) = 𝑔 2 = 𝑔★2 (𝐺 1 ), however, is not affected by the fixed-point an-
(11 − 2𝑥 − 3𝑥𝐺 1 )/(13𝑥 − 34) in the infrared. For 𝑥 below, but nihilation and persists for all 𝑥 > 0. Since the runaway flow
not too far from 11/2, the infrared attractive plane 𝑔 2 = 𝑔★2 (𝐺 1 ) corresponding to this quantum critical fixed point is towards
hosts a fully infrared stable fixed point that can be associated (𝐺 1 , 𝐺 2 ) → (−∞, −∞), one should expect it to describe an in-
with the conformal phase of many-flavor QCD4 , the so-called stability towards a different ground state, spontaneously break-
Caswell-Banks-Zaks fixed point (CBZ) [74, 75]. In addition, ing a symmetry different from chiral symmetry. This suggests
the plane features two fixed points with unique infrared rele- a continuous order-to-order transition for 𝑥 < 𝑥c governed by
vant directions, which can be associated to quantum critical QCP1 .

[1] I. F. Herbut and L. Janssen, Topological Mott Insulator in Three- orders and multipolar Weyl fermions in 227 pyrochlore iridates,
Dimensional Systems with Quadratic Band Touching, Phys. Rev. Phys. Rev. B 103, L241116 (2021).
Lett. 113, 106401 (2014). [6] D. J. Moser and L. Janssen, Quasiuniversality from all-in–all-
[2] L. Janssen and I. F. Herbut, Phase diagram of electronic systems out Weyl quantum criticality in pyrochlore iridates, Phys. Rev.
with quadratic Fermi nodes in 2 < 𝑑 < 4: 2 + 𝜖 expansion, 4 − 𝜖 B 109, L081111 (2024).
expansion, and functional renormalization group, Phys. Rev. B [7] L. Janssen and I. F. Herbut, Nematic quantum criticality in three-
95, 075101 (2017). dimensional Fermi system with quadratic band touching, Phys.
[3] I. Boettcher and I. F. Herbut, Anisotropy induces non- Rev. B 92, 045117 (2015).
Fermi-liquid behavior and nematic magnetic order in three- [8] L. Janssen and I. F. Herbut, Excitonic instability of three-
dimensional Luttinger semimetals, Phys. Rev. B 95, 075149 dimensional gapless semiconductors: Large-𝑁 theory, Phys.
(2017). Rev. B 93, 165109 (2016).
[4] L. Savary, E.-G. Moon, and L. Balents, New Type of Quantum [9] We follow the notation of Ref. [6]. The five real Gell-Mann
Criticality in the Pyrochlore Iridates, Phys. Rev. X 4, 041027 matrices, and therefore also the ℓ = 2 real spherical harmonics
(2014). used in Ref. [7], correspond to the ones used here by relabelling
[5] K. Ladovrechis, T. Meng, and B. Roy, Competing magnetic indices (1, 2, 3, 4, 5) ↦→ (4, 3, 2, 1, 5).
 18

[10] S. Ray and L. Janssen, Gross-Neveu-Heisenberg criticality Lost Conformality in a non-Hermitian Quantum 5-state Potts
from competing nematic and antiferromagnetic orders in bilayer Model, arXiv:2403.00852.
graphene, Phys. Rev. B 104, 045101 (2021). [35] L. Rosa, P. Vitale, and C. Wetterich, Critical Exponents of the
[11] M. Vojta, Y. Zhang, and S. Sachdev, Renormalization group Gross-Neveu Model from the Effective Average Action, Phys.
analysis of quantum critical points in 𝑑-wave superconductors, Rev. Lett. 86, 958 (2001).
Int. J. Mod. Phys. B 14, 3719 (2000). [36] N. Dupuis, L. Canet, A. Eichhorn, W. Metzner, J. Pawlowski,
[12] M. S. Dresselhaus, G. Dresselhaus, and A. Jorio, Group theory: M. Tissier, and N. Wschebor, The nonperturbative functional
application to the physics of condensed matter (Springer Berlin, renormalization group and its applications, Phys. Rep. 910, 1
Heidelberg, 2007). (2021).
[13] H. Gies and C. Wetterich, Renormalization flow of bound states, [37] Y. Huh and S. Sachdev, Renormalization group theory of ne-
Phys. Rev. D 65, 065001 (2002). matic ordering in 𝑑-wave superconductors, Phys. Rev. B 78,
[14] J. M. Pawlowski, Aspects of the functional renormalisation 064512 (2008).
group, Ann. Phys. 322, 2831 (2007). [38] J. Schwab, L. Janssen, K. Sun, Z. Y. Meng, I. F. Herbut, M. Vojta,
[15] S. Floerchinger and C. Wetterich, Exact flow equation for com- and F. F. Assaad, Nematic Quantum Criticality in Dirac Systems,
posite operators, Phys. Lett. B 680, 371 (2009). Phys. Rev. Lett. 128, 157203 (2022).
[16] L. Janssen and H. Gies, Critical behavior of the (2 + 1)- [39] Y.-C. He, M. P. Zaletel, M. Oshikawa, and F. Pollmann, Signa-
dimensional Thirring model, Phys. Rev. D 86, 105007 (2012). tures of Dirac Cones in a DMRG Study of the Kagome Heisen-
[17] A. A. Abrikosov and S. D. Beneslavskii, Possible existence berg Model, Phys. Rev. X 7, 031020 (2017).
of substances intermediate between metals and dielectrics, Zh. [40] Y.-C. He, Y. Fuji, and S. Bhattacharjee, Kagome spin liquid:
Eksp. Teor. Fiz. 59, 1280 (1970), [Sov. Phys. JETP 32, 699 a deconfined critical phase driven by 𝑈 (1) gauge fluctuation,
(1971)]. arXiv:1512.05381.
[18] A. A. Abrikosov, Calculation of critical indices for zero-gap [41] L. Janssen and Y.-C. He, Critical behavior of the QED3 -Gross-
semiconductors, Zh. Eksp. Teor. Fiz. 66, 1443 (1974), [Sov. Neveu model: Duality and deconfined criticality, Phys. Rev. B
Phys. JETP 39, 709 (1974)]. 96, 205113 (2017).
[19] E.-G. Moon, C. Xu, Y. B. Kim, and L. Balents, Non-Fermi- [42] B. Ihrig, L. Janssen, L. N. Mihaila, and M. M. Scherer, De-
Liquid and Topological States with Strong Spin-Orbit Coupling, confined criticality from the QED3 -Gross-Neveu model at three
Phys. Rev. Lett. 111, 206401 (2013). loops, Phys. Rev. B 98, 115163 (2018).
[20] B. I. Halperin, T. C. Lubensky, and S.-k. Ma, First-Order Phase [43] N. Zerf, P. Marquard, R. Boyack, and J. Maciejko, Critical be-
Transitions in Superconductors and Smectic-𝐴 Liquid Crystals, havior of the QED3 -Gross-Neveu-Yukawa model at four loops,
Phys. Rev. Lett. 32, 292 (1974). Phys. Rev. B 98, 165125 (2018).
[21] I. F. Herbut and Z. Tešanović, Critical Fluctuations in Supercon- [44] R. Boyack, A. Rayyan, and J. Maciejko, Deconfined criticality
ductors and the Magnetic Field Penetration Depth, Phys. Rev. in the QED3 Gross-Neveu-Yukawa model: The 1/𝑁 expansion
Lett. 76, 4588 (1996). revisited, Phys. Rev. B 99, 195135 (2019).
[22] I. F. Herbut and Z. Tešanović, Herbut and Tešanović Reply:, [45] E. Dupuis, R. Boyack, and W. Witczak-Krempa, Anomalous
Phys. Rev. Lett. 78, 980 (1997). Dimensions of Monopole Operators at the Transitions between
[23] B. Ihrig, N. Zerf, P. Marquard, I. F. Herbut, and M. M. Scherer, Dirac and Topological Spin Liquids, Phys. Rev. X 12, 031012
Abelian Higgs model at four loops, fixed-point collision, and (2022).
deconfined criticality, Phys. Rev. B 100, 134507 (2019). [46] H. Gies and L. Janssen, UV fixed-point structure of the three-
[24] M. S. Block, R. G. Melko, and R. K. Kaul, Fate of CP 𝑁 −1 Fixed dimensional Thirring model, Phys. Rev. D 82, 085018 (2010).
Points with 𝑞 Monopoles, Phys. Rev. Lett. 111, 137202 (2013). [47] L. Janssen, Spontaneous breaking of Lorentz symmetry in (2 +
[25] M. Song, J. Zhao, L. Janssen, M. M. Scherer, and Z. Y. Meng, 𝜖)-dimensional QED, Phys. Rev. D 94, 094013 (2016).
Deconfined quantum criticality lost, arXiv:2307.02547. [48] J. Braun, H. Gies, L. Janssen, and D. Roscher, Phase structure
[26] A. Nahum, P. Serna, J. T. Chalker, M. Ortuño, and A. M. So- of many-flavor QED3 , Phys. Rev. D 90, 036002 (2014).
moza, Emergent SO(5) Symmetry at the Néel to Valence-Bond- [49] L. Di Pietro, Z. Komargodski, I. Shamir, and E. Stamou, Quan-
Solid Transition, Phys. Rev. Lett. 115, 267203 (2015). tum Electrodynamics in 𝑑 = 3 from the 𝜖 Expansion, Phys. Rev.
[27] T. Senthil, Deconfined quantum critical points: a review, Lett. 116, 131601 (2016).
arXiv:2306.12638. [50] I. F. Herbut, Chiral symmetry breaking in three-dimensional
[28] L. Fei, S. Giombi, and I. R. Klebanov, Critical 𝑂 (𝑁) models in quantum electrodynamics as fixed point annihilation, Phys. Rev.
6 − 𝜖 dimensions, Phys. Rev. D 90, 025018 (2014). D 94, 025036 (2016).
[29] I. F. Herbut and L. Janssen, Critical 𝑂 (2) and 𝑂 (3) 𝜑4 theories [51] H. J. Liao, Z. Y. Xie, J. Chen, Z. Y. Liu, H. D. Xie, R. Z. Huang,
near six dimensions, Phys. Rev. D 93, 085005 (2016). B. Normand, and T. Xiang, Gapless Spin-Liquid Ground State
[30] J. A. Gracey, I. F. Herbut, and D. Roscher, Tensor 𝑂 (𝑁) model in the 𝑆 = 1/2 Kagome Antiferromagnet, Phys. Rev. Lett. 118,
near six dimensions: Fixed points and conformal windows from 137202 (2017).
four loops, Phys. Rev. D 98, 096014 (2018). [52] X. Chen, S.-J. Ran, T. Liu, C. Peng, Y.-Z. Huang, and G. Su,
[31] V. Gorbenko, S. Rychkov, and B. Zan, Walking, Weak first-order Thermodynamics of spin-1/2 Kagomé Heisenberg antiferro-
transitions, and Complex CFTs II. Two-dimensional Potts model magnet: algebraic paramagnetic liquid and finite-temperature
at 𝑄 > 4, SciPost Phys. 5, 050 (2018). phase diagram, Sci. Bull. 63, 1545 (2018).
[32] H. Ma and Y.-C. He, Shadow of complex fixed point: Approx- [53] M. Hermele, T. Senthil, and M. P. A. Fisher, Algebraic spin
imate conformality of 𝑄 > 4 Potts model, Phys. Rev. B 99, liquid as the mother of many competing orders, Phys. Rev. B 72,
195130 (2019). 104404 (2005).
[33] J. L. Jacobsen and K. J. Wiese, Lattice Realization of Complex [54] V. P. Gusynin and P. K. Pyatkovskiy, Critical number of fermions
Conformal Field Theories: Two-Dimensional Potts Model with in three-dimensional QED, Phys. Rev. D 94, 125009 (2016).
𝑄 > 4 States, Phys. Rev. Lett. 133, 077101 (2024). [55] A. V. Kotikov and S. Teber, Critical behavior of (2 + 1)-
[34] Y. Tang, H. Ma, Q. Tang, Y.-C. He, and W. Zhu, Reclaiming the dimensional QED: 1/𝑁 𝑓 corrections in an arbitrary nonlocal
 19

gauge, Phys. Rev. D 94, 114011 (2016). Phys. Rev. Lett. 130, 186701 (2023).
[56] N. Karthik and R. Narayanan, No evidence for bilinear conden- [66] A. Nahum, Fixed point annihilation for a spin in a fluctuating
sate in parity-invariant three-dimensional QED with massless field, Phys. Rev. B 106, L081109 (2022).
fermions, Phys. Rev. D 93, 045020 (2016). [67] G. Cuomo, Z. Komargodski, M. Mezei, and A. Raviv-Moshe,
[57] N. Karthik and R. Narayanan, Scale invariance of parity- Spin impurities, Wilson lines and semiclassics, J. High Energy
invariant three-dimensional QED, Phys. Rev. D 94, 065026 Phys. 06 (2022) 112.
(2016). [68] H. Gies, J. Jaeckel, and C. Wetterich, Towards a renormalizable
[58] A. M. Sengupta, Spin in a fluctuating field: The Bose(+Fermi) standard model without a fundamental Higgs scalar, Phys. Rev.
Kondo models, Phys. Rev. B 61, 4041 (2000). D 69, 105008 (2004).
[59] S. Sachdev, C. Buragohain, and M. Vojta, Quantum Impurity in [69] H. Gies and J. Jaeckel, Chiral phase structure of QCD with many
a Nearly Critical Two-Dimensional Antiferromagnet, Science flavors, Eur. Phys. J. C 46, 433 (2006).
286, 2479 (1999). [70] J. Braun, C. S. Fischer, and H. Gies, Beyond Miransky scaling,
[60] S. Sachdev, Quantum impurity in a magnetic environment, J. Phys. Rev. D 84, 034045 (2011).
Stat. Phys. 115, 47 (2004). [71] Y. Kusafuka and H. Terao, Fixed point merger in the 𝑆𝑈 (𝑁)
[61] M. Vojta, Impurity quantum phase transitions, Philos. Mag. 86, gauge beta functions, Phys. Rev. D 84, 125006 (2011).
1807 (2006). [72] S. Gukov, RG flows and bifurcations, Nucl. Phys. B 919, 583
[62] L. Zhu and Q. Si, Critical local-moment fluctuations in the (2017).
Bose-Fermi Kondo model, Phys. Rev. B 66, 024426 (2002). [73] F. Kuipers, U. Gürsoy, and Y. Kuznetsov, Bifurcations in the
[63] G. Zaránd and E. Demler, Quantum phase transitions in the RG-flow of QCD, J. High Energy Phys. 07 (2019) 075.
Bose-Fermi Kondo model, Phys. Rev. B 66, 024427 (2002). [74] W. E. Caswell, Asymptotic Behavior of Non-Abelian Gauge
[64] M. Weber, Tunable quantum criticality and pseudocriticality Theories to Two-Loop Order, Phys. Rev. Lett. 33, 244 (1974).
across the fixed-point annihilation in the anisotropic spin-boson [75] T. Banks and A. Zaks, On the phase structure of vector-like
model, arXiv:2403.02400. gauge theories with massless fermions, Nucl. Phys. B 196, 189
[65] M. Weber and M. Vojta, SU(2)-Symmetric Spin-Boson Model: (1982).
Quantum Criticality, Fixed-Point Annihilation, and Duality, [76] D. B. Kaplan, J.-W. Lee, D. T. Son, and M. A. Stephanov,
Conformality lost, Phys. Rev. D 80, 125005 (2009).