Entropy spectroscopy of a bilayer graphene quantum dot

C. Adam,1, ∗ H. Duprez,1 N. Lehmann,1 A. Yglesias,1 S. Cances,1 M. J. Ruckriegel,1 M. Masseroni,1 C. Tong,1 A. O.

Denisov,1 W. Huang,1 D. Kealhofer,1 R. Garreis,1 K. Watanabe,2 T. Taniguchi,3 K. Ensslin,1 and T. Ihn1
1
Solid State Physics Laboratory, ETH Zurich, Zurich CH-8093, Switzerland
2
Research Center for Electronic and Optical Materials,
National Institute for Materials Science, 1-1 Namiki, Tsukuba 305-0044, Japan
3
Research Center for Materials Nanoarchitectonics,
National Institute for Materials Science, 1-1 Namiki, Tsukuba 305-0044, Japan
(Dated: December 25, 2024)
arXiv:2412.18000v1 [cond-mat.mes-hall] 23 Dec 2024

We measure the entropy change of charge transitions in an electrostatically defined quantum

dot in bilayer graphene. Entropy provides insights into the equilibrium thermodynamic properties
of both ground and excited states beyond transport measurements. For the one-carrier regime,
the obtained entropy shows that the ground state has a two-fold degeneracy lifted by an out-of-
plane magnetic field. This observation is in agreement with previous direct transport measurements
and confirms the applicability of this novel method. For the two-carrier regime, the extracted
entropy indicates a non-degenerate ground state at zero magnetic field, contrary to previous studies
suggesting a three-fold degeneracy. We attribute the degeneracy lifting to the effect of Kane-Mele
type spin–orbit interaction on the two-carrier ground state, which has not been observed before. Our
work demonstrates the validity and efficacy of entropy measurements as a unique, supplementary
experimental tool to investigate the degeneracy of the ground state in quantum devices build in
materials such as graphene. This technique, applied to exotic systems with fractional ground state
entropies, will be a powerful tool in the study of quantum matter.

I. INTRODUCTION graphene systems as they often involve highly specialized

measurement techniques [20, 21]. On the other hand, the
The thermodynamic entropy of a system is used to entropy-to-charge conversion, pioneered in ref. [8], which
study the degeneracy of the quantum mechanical ground requires to controllably define a charge island (with a
state of exotic phases of condensed matter. The non- charging energy substantially larger than the thermal en-
abelian fractional quantum Hall state at filling fac- ergy) connected to a thermal bath, is accessible through
tor 5/2 [1–3], Majorana fermions [4] and multi-channel standard transport measurements across a charge detec-
Kondo systems [5] are predicted to exhibit fractional en- tor. This scheme is particularly suited to rhombohedral
tropy specific to the respective quasiparticles constitut- graphene systems in which gate-defined devices can be
ing the phase. For these highly engineered systems, ac- formed thanks to the bandgap opening with a transverse
cessing the entropy with conventional methods such as displacement field [22–25]. Here, we demonstrate such
used for bulk systems, e.g. measurements of heat capac- capacitance-based entropy measurements in a dual-gated
ity, is experimentally challenging [6]. Therefore, indirect (rhombohedral or Bernal) bilayer graphene (BLG) quan-
schemes have been implemented to measure the entropy tum dot, an object typically studied for semiconduct-
of mesoscopic systems, such as extracting the entropy ing qubit purposes [26–28]. We measured the entropy
from thermoelectric transport measurements [7]. An ap- change upon changing the quantum dot’s charge state.
proach exploiting thermodynamic Maxwell relations [8, 9] For a weakly-coupled dot to the reservoir, the entropy is
has already been experimentally demonstrated for a lo- directly connected to the charge ground state degener-
calized spin-1/2 electron in an electrostatically defined acy. Notably, we measured that the ground state degen-
quantum dot (QD) in a (Al)GaAs heterostructure. It eracy of the second charge carrier is lower than expected,
has been further developed to measure the entropy of a supported by excited state spectroscopy. We propose a
Kondo spin-singlet [10, 11], and another thermodynamic model to explain this surprising behavior.
quantity: the analogue to the magnetic susceptibility of
a spin impurity in a ’charge’ Kondo circuit [12].
Extending this technique to graphene multilayer sys- II. DEVICE & EXPERIMENTAL SETUP
tems is particularly attractive, as these have been shown
to host a variety of exotic electronic phases, such as corre- The schematic of the system is shown in Fig. 1(a). It
lated insulators, the anomalous quantum Hall effect, and consists of a QD in thermal equilibrium with a bath of
superconductivity [13–19]. However, only few studies re- hole carriers (“reservoir”), and a charge detection circuit
port on measurements of thermodynamic quantities in (“detector”). A current Iheater is driven through electro-
statically defined constrictions called heaters. The ohmic
resistance of the heaters causes power to be dissipated as
heat. This induces a temperature change of the reservoir
∗ adamc@phys.ethz.ch due to Joule heating [29].
 2

(a) (b) RB
PG QD
VPG LB
Vheater,1
Vdet detec-
tor
reservoir
SGM AlOx
top SGB
hBN
Idet bottom BLG
QD
Iheater hBN
Vheater,2 back
gate

SG FG contact

(c) (f)

(d) (g)

(e) (h)

FIG. 1. (a) The sample consists of a quantum dot (QD) thermally coupled to a reservoir. The reservoir temperature T (t) is
changed by driving a current Iheater through a heater structure defined by split gates (SG) and finger gates (FG). The charge
detector (CD), capacitively coupled to the QD, carries a current Idet which changes as the number of charge carriers changes
on the QD. (b) Schematic of the van der Waals (vdW) stack in the region of the QD: bilayer graphene (BLG) is encapsulated
between two layers of hexagonal boron nitride (hBN) as dielectric. To form a conduction channel, the graphite back gate, BG,
and gold split gates, SGM and SGB, are used to open a band gap and tune the hole carrier density in the BLG to deplete it of
charge carriers underneath the gates. The finger gates, LB, PG, RB, are separated from the split gates by an aluminum oxide
layer (AlOx) and used to define the tunnel barriers and tune the QD. For the 0 → 1- and 1 → 2-transition, respectively: The
DC component of the detector current (c) and (f) with the extracted temperature modulation (continuous traces are fits to
eq. 5), the to I0 normalized second harmonic current (d) and (g) with extracted entropy change (continuous traces are fits to
eq.6), and the entropy obtained by integration (e) and (h) (continuous traces are plots of eq. D14 in (e) and eq. D18 in (h)).
The gray bars ±0.1 ln(2) as a guide to the eye.

Fig. 1(b) shows a schematic side view of the van in the hole conduction regime. The barriers are tuned
der Waals (vdW) stack in which the QD is defined: such that there is no coupling to the left lead. A more
BLG is encapsulated by hexagonal boron nitride (hBN) detailed description of the sample and its operation can
surrounded by a global graphite back gate (BG) and be found in Appendix A.
with two layers of gold top gates: split gates SGM, SGB
to form a one-dimensional channel, and finger gates to
form the QD (plunger gate PG and tunneling barrier
gates LB, RB). The top gate layers are separated by III. ENTROPY EXTRACTION
aluminum oxide (AlOx) as a dielectric. The displace-
ment field is roughly −0.46 V/nm resulting in a band The relevant thermodynamic state variables of the
gap of approximately 50 meV[30]. The device is operated QD–reservoir system include the mean occupation num-
 3

ber N and entropy S of the QD, as well as the electro- QD is then described by
chemical potential µ and carrier temperature T in the
reservoir. In our experiment, µ and T are independent, 1
N (µ, T ) = + N − 1, (7)
externally controlled variables. In an experiment con- µN −µ−T ∆S N −1→N

ducted at constant temperature T , these variables are 1+e kB T

connected through the Maxwell relation   µN −µ− ∆S N −1→N

∂N 4kB T
2
= . (8)
ˆ

∂T cosh2 µN −µ−T ∆S N −1→N
µ
∂N (µ, T ) µ,T =T
 
S(µ, T ) − S(µ0 , T ) = dµ. (1) 2kB T
µ0 ∂T µ
where µN − µ = αPG e(VPG − V0 ) and µN is the elec-
trochemical potential of the quantum dot for adding the
By measuring (∂N /∂T )µ one can get the entropy of the
N -th charge carrier, αPG ≈ 0.025 is the plunger gate
system. In order to measure this quantity, we control the
lever arm, and V0 is a plunger gate voltage offset. Note
temperature T (t) by driving an electrical current
that changing µ in eqs. 7, 8 is equivalent to changing the
energy difference µN − µ, which is accomplished in the
Iheater (t) = Iheater · cos(ωt) (2) experiment by changing VPG . The measurements for ob-
taining αPG are shown in Appendix C. Eq. 8 contains
through the heaters. The temperature resulting from ∆S N −1→N as a parameter. Since Idet2ω
∝ (∂N /∂T )µ,T =T
Joule heating can be expressed as [29] one can simply fit eq. 8 to Idet and obtain ∆S N −1→N .
2ω

q This can be done without performing a temperature cal-

T (t) = T02 + aIheater
2 (t) (3) ibration.
Method B: Eq. 6 allows to get (∂N /∂T )µ,T =T from
≈ T0 + ∆T + ∆T cos(2ωt), (4) 2ω
Idet , without assumptions about its functional shape.
However, I0 and ∆T have to be known. I0 can be read
where T0 is the reservoir temperature without any heat- from Idet
DC
as shown in Fig. 1(c),(f). ∆T is set by the
ing current and a is a system-specific constant which is heating current Iheater and calibrated in the beginning.
determined by a calibration measurement. The average After obtaining (∂N /∂T )µ,T =T one can get the entropy
temperature increases to T = T0 + ∆T , while a tempera-
by using eq. 1. The integration along µ is equivalent to
ture modulation ∆T = aIheater
2
/4T0 with frequency 2ω is
an integration along VPG , knowing αPG .
introduced. The frequency ω is chosen sufficiently low to
ensure that the QD is in equilibrium with the reservoir at
all times. The charge detector (CD) capacitively coupled
to the QD [31] allows us to extract the mean occupation IV. ONE- & TWO-CARRIER ENTROPY
number N via the detector current
Fig. 1(c)-(e) show the measured charge detector cur-
I DC
det rent and extracted entropy for the transition between
the 0 → 1 charge state. The DC component of the de-
z }| {
Idet (t) ≈ I0 · N (µ, T ) + Ioff + γVPG (5)
  tector current Idet
DC
and the second harmonic component
∂N 2ω
Idet are measured with a voltmeter and lock-in ampli-
+ I0 · ∆T · cos(2ωt) (6)
∂T µ,T =T fier attached to a transimpedance amplifier. The result-
| {z
I 2ω
} ing traces are obtained by scanning the 0 → 1 transi-
det
tion several times and averaging the individual traces.
The changing plunger gate VPG corresponds to an ef-
where I0 is the change in CD current in response to a fective change of the electrochemical potential µ of the
charge change on the QD, Ioff the offset current through reservoir by ∆µ = −αPG e(VPG − V0 ). As the vacuum
the CD at the operation point, VPG the plunger gate state |0⟩ is non-degenerate, we define the entropy to be
voltage used to change the QD occupation, and γ the S(N = 0) = 0. Gradually changing the hole occupa-
transconductance of the CD with respect to the plunger tion number to N = 1 by lowering VPG , the entropy
gate. The DC component Idet DC
is used to determine reaches a maximum of kB ln(3) before settling at kB ln(2).
I0 , while the second harmonic component Idet2ω
provides The entropy change for the 0 → 1 transition is therefore
(∂N /∂T )µ,T =T , once ∆T has been determined by cali- ∆S 0→1 = kB ln(2). This can be understood based on
bration. The entropy difference ∆S N −1→N between the the energy level structure of the one-carrier charge state:
integer charge states N − 1 and N is obtained in two Its ground state is the Kramers pair |K − ↓⟩, |K + ↑⟩,
different ways. hence two-fold degenerate at zero magnetic field. The re-
Method A: In the thermally broadened regime, i.e. sult confirms observations recently made by performing
kB T (t) ≫ ΓR with ΓR characterizing the tunnel broad- counting statistics measurements on BLG QDs [32]. The
ening, the QD occupation close to a transition between entropy peak at kB ln(3) corresponds to the QD being
integer numbers N − 1 and N of charge carriers in the equally likely in any of the following three microstates:
 4

for N = 0 this is the empty state |0⟩, and for N = 1 this broadened regime and fitting eq.5 to Idet
DC
. The evolution
is the two-fold degenerate Kramers pair |K − ↓⟩, |K + ↑⟩. of T in Iheater follows eq. 3 as can be seen from the
Figures 1(f)-(g) show the measured results for the fit. The experiments for the 0 → 1 and 1 → 2 transi-
1 → 2 charge carrier transition. Starting out at N = 1 tion were carried out at different cryostat temperatures
with the initial entropy kB ln(2), gradually changing the T0 . The data is well described by eq. 3. From Fig. 2 (a)
occupation number to N = 2 by further lowering VPG , the temperature modulation ∆T = T − T0 is extracted
the entropy has a maximum with kB ln(3) before settling for various heater currents. Fig. 2 (b) shows the entropy
at S(N = 2) = 0. The entropy change for the 1 → 2 changes ∆S 0→1 , ∆S 1→2 obtained by applying method
transition is therefore ∆S 1→2 = −kB ln(2). We con- A and B. Both methods show reasonable agreement with
clude from this result that the ground state for N = 2 each other.
particles in the QD is non-degenerate. This result is sur-
prising as previous devices studied in higher magnetic
fields suggested a valley-singlet spin-triplet ground state V. ENTROPY IN MAGNETIC FIELD
with a three-fold spin-degeneracy, a conclusion reached
by extrapolating finite bias measurements to zero mag- Figures 3(a) and (b) show the entropy changes ∆S A 0→1
netic field [33, 34]. In our experiment, the entropy peak and ∆S A 1→2 for the 0 → 1 and 1 → 2 transition, respec-
at kB ln(3) corresponds to the QD being equally likely tively, as a function of magnetic field B⊥ . In Fig 3(a),
in any of the microstates of N = 1 (two microstates, upon increasing B⊥ the entropy change decreases from
Kramers pair) and N = 2 (only one microstate), corre- kB ln(2) to 0. This is in agreement with the two-fold
sponding in total to three microstates of equal probability degenerate ground ground state at zero magnetic field
1/3. In Sec. V we study the change of entropy in mag- given by the Kramers pair |K − ↓⟩, |K + ↑⟩. The degen-
netic field and from that provide a possible explanation eracy is gradually lifted with increasing B⊥ due to the
for the lifting of degeneracy at zero magnetic field for the spin- and valley-Zeeman effect: Once the energy split-
second charge carrier state. ting gv µB B⊥ between states |K − ↓⟩ and |K + ↑⟩ exceeds
Fig. 2 (a) shows the mean temperature T of the hole the thermal energy kB T , the degeneracy is fully lifted
bath forming the reservoir as a function of the heater in the thermodynamic sense, and the unambiguous ther-
current Iheater . Temperatures for individual heater cur- modynamic ground state is |K − ↓⟩. Similar behaviour
rents are obtained by tuning the QD into the thermally is also observed for the 1 → 2-transition in Fig 3(b) at
magnetic fields below 100 mT: upon increasing B⊥ the
entropy change increases from −kB ln(2) to 0 as the de-
(a) generacy of the one carrier state is lifted. Further increas-
ing B⊥ leads to a peak in entropy change of kB ln(2) at
210 mT. Since here, the one-carrier ground state degen-
eracy is fully lifted already, the change in entropy must be
fully due to a ground state degeneracy in the two-carrier
charge state.
We compare the extracted entropy change in magnetic
field with the energy level spectrum of the QD obtained
from finite bias spectroscopy measurements performed in
an out-of-plane magnetic which are shown in Fig. 3(c)-
(b) (d). For that we operate the QD into a regime where it
is symmetrically coupled to two leads by adjusting the
voltages of the barrier gates LB and RB. This is different
from the regime in which the entropy measurements are
performed, where the QD is only coupled to one lead,
with the coupling controlled by RB. A source-drain bias
VSD is applied as well as a voltage modulation to the
plunger gate PG in order to measure the transconduc-
tance ∂Idet /∂VPG . By carefully tuning the coupling to
the leads, an excited state in the bias window increases
the average occupation number of the dot, which ex-
FIG. 2. (a) Average temperatures T of the holes tunnel-
ing onto the QD as a function of the heater current Iheater
presses itself in the appearance of a peak in the transcon-
for the 0 → 1- and 1 → 2-transition, respectively. The solid ductance. These peaks are labeled in Fig. 3 and identi-
lines show a fit with eq. 3. The temperature at zero heating fied according to the g-factors extracted. In Fig. 3(c) this
currents differ as the experiments where carried out at differ- is the Kramers pair |K − ↓⟩, |K + ↑⟩, with the g-factors
ent cryostat base temperatures. (b) Entropy changes for the (1)
for the valley- and spin-Zeeman effects gv = 14 and
0 → 1 and 1 → 2 transitions obtained using Method A and gs = 2, respectively, for the first hole carrier in the QD
B. The results of both methods are in reasonable agreement. as well as the Kane-Mele spin–orbit gap ∆SO = 75 µeV.
 5

(a) (b)

(c) (d)

FIG. 3. Evolution of the measured entropy change in out-of-plane magnetic field together with the calculated entropy change
(solid line) for the 0 → 1 (a) and 1 → 2 transition (b). for the 0 → 1 and 1 → 2 transitions. Finite bias spectroscopy
measurements in out-of-plane magnetic field performed for the 0 → 1 (c) and 1 → 2 transition (d). The states are identified by
there behaviour in out-of-plane magnetic field.

In Fig. 3(d) the ground state is labeled as |Sv Ts0 ⟩⊕|Tv0 Ss ⟩. 0 → 1 transition is consistent with the excited state spec-
Its energy does not change in out-of-plane magnetic field trum measured with finite bias spectroscopy and also
as there is no valley- or spin-Zeeman effect, as discussed previous findings [33, 34]: The ground state has a two-
in Sec. VI. The shift of the transconductance peak in fold degeneracy (Kramers pair |K − ↓⟩, |K + ↑⟩) lifted in
magnetic field comes from the fact that the addition en- out-of-plane magnetic field due to the valley- and spin-
ergy for the second carrier is affected by the energy of Zeeman effect. For the 1 → 2 transition the extracted
the first carrier. We observe a state crossing at 210 mT entropy change reveals a non-degenerate ground state for
caused by a valley-triplet spin-singlet state labeled as the two-carrier charge state at zero magnetic field. This
|Tv− Ss ⟩. Its energy gap to the ground state at zero- has not been observed before. Previous studies hinted
magnetic field is ∆ = 205 µeV. From the shift of the towards a three-fold degenerate ground state given by a
transconductance peak of |Sv Ts0 ⟩ ⊕ |Tv0 Ss ⟩ and |Tv− Ss ⟩, valley-singlet spin-triplet states [33, 34].
the g-factor of the valley-Zeeman effect in the two carrier We attribute this finding to the effect of the Kane-Mele
(2)
regime is found to be gv = 16, supporting the proposed type spin–orbit interaction [36, 37] present in BLG and
energy spectrum discussed in Sec. VI. Their values also the exceptionally small exchange splitting between dif-
agree with previous findings [33–35]. ferent valley states. A model is developed (Appendix B)
In the thermally broadened limit, the entropy evolu- with the resulting energy spectrum shown in Fig. 4.
tion in magnetic field can be calculated. The calculation For the two-carrier charge state, the Kane-Mele type
is carried out in Appendix D and results in expressions spin–orbit interaction leads to a coupling of the valley-
for ∆S calc calc
0→1 (eq. D32) and ∆S 1→2 (eq. D33). The corre-
singlet spin-triplet state |Sv Ts0 ⟩ with the valley-triplet
sponding traces are plotted in Figs. 3 (c) and (d). They spin-singlet state |Tv0 Ss ⟩. This results in the three-fold
show reasonable agreement with the measured entropy degeneracy being lifted. The new ground state is denoted
changes. as |Sv Ts0 ⟩ ⊕ |Tv0 Ss ⟩. Its energy is lowered by
s 2
J1 + J2 J1 + J2
∆′SO = + ∆2SO − (9)
VI. DISCUSSION & OUTLOOK 2 2

Measuring entropy allows us to directly determine the compared to the valley-singlet spin-triplet states |Sv Ts− ⟩,
ground state degeneracy of charge states in a BLG QD. |Sv Ts+ ⟩. To give an estimate, we assume that J1 ≈ J2
This approach gives insights beyond conventional trans- and set J1 = J2 = ∆ = 205 µeV and ∆SO = 75 µeV. Pre-
port spectroscopy techniques where degeneracies are de- vious experiments showed J1 , J2 ≃ 1000 µeV [33]. This
tected in an indirect way by the energy levels splitting results in ∆′SO = 25 µeV. The extended data set on fi-
in magnetic field. The extracted entropy change for the nite bias spectroscopy presented in Appendix C shows
 6

the presence of a gap between ground state and first ex-

cited state of roughly 45 µeV. Furthermore, the energy
of the first excited state in the two-carrier charge state
shifts with a g-factor of 2, suggesting that it is indeed a
spin-triplet state. The fact that entropy spectroscopy al-
lows us to detect any residual degeneracies that are not
lifted by magnetic field and that the resolution in en-
ergy is only limited by kB T ≈ 8 µeV make this method a
valuable addition to conventional transport spectroscopy
methods.
Our work generalizes and extends the Maxwell
relation-based entropy measurements to samples fabri-
cated in novel 2D vdW heterostructures and shows that
sophisticated device designs can be implemented in these
material systems, paving the way to probe various ex-
otic states in BLG devices. Examples of these are the
(multichannel-) Kondo effect [38, 39] or quantum spin
chains with fractional or integer spins [40, 41].

ACKNOWLEDGMENTS

We appreciate the many discussions with Josh Folk,

Eran Sela, and Yigal Meir concerning the concept of en-
tropy in mesoscopic systems. We thank Peter Märki,
Thomas Bähler, as well as the FIRST staff for their tech-
nical support. We acknowledge support from the Euro-
pean Graphene Flagship Core3 Project, Swiss National
Science Foundation via NCCR Quantum Science, and
FIG. 4. Excited state spectra for the one-carrier and two- H2020 European Research Council (ERC) Synergy Grant
carrier state in an out-of-plane magnetic field including the under Grant Agreement 95154.
effect of Kane-Mele type spin–orbit interaction for the N = 2
charge state. The ground state |Sv Ts0 ⟩ ⊕ |Tv0 Ss ⟩ is separated K.W. and T.T. acknowledge support from the JSPS
from the spin-triplet states |Sv Ts+ ⟩, |Sv Ts− ⟩ by an energy gap KAKENHI (Grant Numbers 20H00354 and 23H02052)
∆′SO . The ground state crossing is due to the valley-triplet and World Premier International Research Center
state |Tv− Ss ⟩ and appears at an out-of-plane magnetic field of Initiative (WPI), MEXT, Japan.
B×.

[1] N. R. Cooper and A. Stern, Observable Bulk Signatures Charge Relations, Physical Review Letters 128, 146803
of Non-Abelian Quantum Hall States, Physical Review (2022).
Letters 102, 176807 (2009). [6] B. A. Schmidt, K. Bennaceur, S. Gaucher, G. Gervais,
[2] G. Ben-Shach, C. R. Laumann, I. Neder, A. Yacoby, and L. N. Pfeiffer, and K. W. West, Specific heat and entropy
B. I. Halperin, Detecting Non-Abelian Anyons by Charg- of fractional quantum Hall states in the second Landau
ing Spectroscopy, Physical Review Letters 110, 106805 level, Physical Review B 95, 201306 (2017).
(2013). [7] Y. Kleeorin, H. Thierschmann, H. Buhmann, A. Georges,
[3] G. Viola, S. Das, E. Grosfeld, and A. Stern, Thermo- L. W. Molenkamp, and Y. Meir, How to measure the en-
electric Probe for Neutral Edge Modes in the Fractional tropy of a mesoscopic system via thermoelectric trans-
Quantum Hall Regime, Physical Review Letters 109, port, Nature Communications 10, 5801 (2019).
146801 (2012). [8] N. Hartman, C. Olsen, S. Lüscher, M. Samani, S. Fallahi,
[4] S. Smirnov, Majorana tunneling entropy, Physical Re- G. C. Gardner, M. Manfra, and J. Folk, Direct entropy
view B 92, 195312 (2015). measurement in a mesoscopic quantum system, Nature
[5] C. Han, Z. Iftikhar, Y. Kleeorin, A. Anthore, F. Pierre, Physics 14, 1083 (2018).
Y. Meir, A. K. Mitchell, and E. Sela, Fractional Entropy [9] E. Sela, Y. Oreg, S. Plugge, N. Hartman, S. Lüscher,
of Multichannel Kondo Systems from Conductance- and J. Folk, Detecting the Universal Fractional Entropy
 7

of Majorana Zero Modes, Physical Review Letters 123, Direct observation of a widely tunable bandgap in bilayer
147702 (2019). graphene, Nature 459, 820 (2009).
[10] T. Child, O. Sheekey, S. Lüscher, S. Fallahi, G. C. [23] C. H. Lui, Z. Li, K. F. Mak, E. Cappelluti, and T. F.
Gardner, M. Manfra, A. Mitchell, E. Sela, Y. Kleeorin, Heinz, Observation of an electrically tunable band gap
Y. Meir, and J. Folk, Entropy Measurement of a Strongly in trilayer graphene, Nature Physics 7, 944 (2011).
Coupled Quantum Dot, Physical Review Letters 129, [24] E. McCann and M. Koshino, The electronic properties
227702 (2022). of bilayer graphene, Reports on Progress in Physics 76,
[11] T. Child, O. Sheekey, S. Lüscher, S. Fallahi, G. C. Gard- 056503 (2013).
ner, M. Manfra, and J. Folk, A Robust Protocol for En- [25] F. Zhang, B. Sahu, H. Min, and A. H. MacDonald, Band
tropy Measurement in Mesoscopic Circuits, Entropy 24, structure of $ABC$-stacked graphene trilayers, Physical
417 (2022). Review B 82, 035409 (2010).
[12] C. Piquard, P. Glidic, C. Han, A. Aassime, A. Cavanna, [26] K. Hecker, L. Banszerus, A. Schäpers, S. Möller, A. Pe-
U. Gennser, Y. Meir, E. Sela, A. Anthore, and F. Pierre, ters, E. Icking, K. Watanabe, T. Taniguchi, C. Volk, and
Observing the universal screening of a Kondo impurity, C. Stampfer, Coherent charge oscillations in a bilayer
Nature Communications 14, 7263 (2023). graphene double quantum dot, Nature Communications
[13] Y. Cao, V. Fatemi, A. Demir, S. Fang, S. L. Tomarken, 14, 7911 (2023).
J. Y. Luo, J. D. Sanchez-Yamagishi, K. Watanabe, [27] R. Garreis, C. Tong, J. Terle, M. J. Ruckriegel, J. D. Ger-
T. Taniguchi, E. Kaxiras, R. C. Ashoori, and P. Jarillo- ber, L. M. Gächter, K. Watanabe, T. Taniguchi, T. Ihn,
Herrero, Correlated insulator behaviour at half-filling K. Ensslin, and W. W. Huang, Long-lived valley states
in magic-angle graphene superlattices, Nature 556, 80 in bilayer graphene quantum dots, Nature Physics , 1
(2018). (2024).
[14] M. Serlin, C. L. Tschirhart, H. Polshyn, Y. Zhang, J. Zhu, [28] A. O. Denisov, V. Reckova, S. Cances, M. J. Ruckriegel,
K. Watanabe, T. Taniguchi, L. Balents, and A. F. Young, M. Masseroni, C. Adam, C. Tong, J. D. Gerber, W. W.
Intrinsic quantized anomalous Hall effect in a moiré het- Huang, K. Watanabe, T. Taniguchi, T. Ihn, K. Ensslin,
erostructure, Science 367, 900 (2020). and H. Duprez, Ultra-long relaxation of a Kramers qubit
[15] Y. Cao, V. Fatemi, S. Fang, K. Watanabe, T. Taniguchi, formed in a bilayer graphene quantum dot (2024).
E. Kaxiras, and P. Jarillo-Herrero, Unconventional super- [29] B. Huard, H. Pothier, D. Esteve, and K. E. Nagaev, Elec-
conductivity in magic-angle graphene superlattices, Na- tron heating in metallic resistors at sub-Kelvin tempera-
ture 556, 43 (2018). ture, Physical Review B 76, 165426 (2007).
[16] Y. Cao, D. Chowdhury, D. Rodan-Legrain, O. Rubies- [30] E. Icking, L. Banszerus, F. Wörtche, F. Volmer,
Bigorda, K. Watanabe, T. Taniguchi, T. Senthil, P. Schmidt, C. Steiner, S. Engels, J. Hesselmann,
and P. Jarillo-Herrero, Strange Metal in Magic-Angle M. Goldsche, K. Watanabe, T. Taniguchi, C. Volk,
Graphene with near Planckian Dissipation, Physical Re- B. Beschoten, and C. Stampfer, Transport Spectroscopy
view Letters 124, 076801 (2020). of Ultraclean Tunable Band Gaps in Bilayer Graphene,
[17] A. M. Seiler, F. R. Geisenhof, F. Winterer, K. Watanabe, Advanced Electronic Materials 8, 2200510 (2022).
T. Taniguchi, T. Xu, F. Zhang, and R. T. Weitz, Quan- [31] A. Kurzmann, H. Overweg, M. Eich, A. Pally, P. Rick-
tum cascade of correlated phases in trigonally warped haus, R. Pisoni, Y. Lee, K. Watanabe, T. Taniguchi,
bilayer graphene, Nature 608, 298 (2022). T. Ihn, and K. Ensslin, Charge Detection in Gate-Defined
[18] Y. Zhang, R. Polski, A. Thomson, E. Lantagne- Bilayer Graphene Quantum Dots, Nano Letters 19, 5216
Hurtubise, C. Lewandowski, H. Zhou, K. Watan- (2019).
abe, T. Taniguchi, J. Alicea, and S. Nadj-Perge, En- [32] H. Duprez, S. Cances, A. Omahen, M. Masseroni, M. J.
hanced superconductivity in spin-orbit proximitized bi- Ruckriegel, C. Adam, C. Tong, R. Garreis, J. D. Ger-
layer graphene, Nature 613, 268 (2023). ber, W. Huang, L. Gächter, K. Watanabe, T. Taniguchi,
[19] T. Han, Z. Lu, Y. Yao, J. Yang, J. Seo, C. Yoon, T. Ihn, and K. Ensslin, Spin-valley locked excited states
K. Watanabe, T. Taniguchi, L. Fu, F. Zhang, and spectroscoy in a one-particle bilayer graphene quantum
L. Ju, Large quantum anomalous Hall effect in spin-orbit dot, Nature Communications 15, 9717 (2024).
proximitized rhombohedral graphene, Science 384, 647 [33] A. Kurzmann, M. Eich, H. Overweg, M. Mangold, F. Her-
(2024). man, P. Rickhaus, R. Pisoni, Y. Lee, R. Garreis, C. Tong,
[20] J. Vallejo Bustamante, N. J. Wu, C. Fermon, K. Watanabe, T. Taniguchi, K. Ensslin, and T. Ihn, Ex-
M. Pannetier-Lecoeur, T. Wakamura, K. Watanabe, cited States in Bilayer Graphene Quantum Dots, Physical
T. Taniguchi, T. Pellegrin, A. Bernard, S. Daddinounou, Review Letters 123, 026803 (2019).
V. Bouchiat, S. Guéron, M. Ferrier, G. Montambaux, [34] S. Möller, L. Banszerus, A. Knothe, C. Steiner, E. Icking,
and H. Bouchiat, Detection of graphene’s divergent or- S. Trellenkamp, F. Lentz, K. Watanabe, T. Taniguchi,
bital diamagnetism at the Dirac point, Science 374, 1399 L. Glazman, V. Fal’ko, C. Volk, and C. Stampfer, Prob-
(2021). ing Two-Electron Multiplets in Bilayer Graphene Quan-
[21] T. Arp, O. Sheekey, H. Zhou, C. L. Tschirhart, C. L. Pat- tum Dots, Physical Review Letters 127, 256802 (2021).
terson, H. M. Yoo, L. Holleis, E. Redekop, G. Babikyan, [35] C. Tong, R. Garreis, A. Knothe, M. Eich, A. Sacchi,
T. Xie, J. Xiao, Y. Vituri, T. Holder, T. Taniguchi, K. Watanabe, T. Taniguchi, V. Fal’ko, T. Ihn, K. Ensslin,
K. Watanabe, M. E. Huber, E. Berg, and A. F. Young, and A. Kurzmann, Tunable Valley Splitting and Bipolar
Intervalley coherence and intrinsic spin–orbit coupling in Operation in Graphene Quantum Dots, Nano Letters 21,
rhombohedral trilayer graphene, Nature Physics 20, 1413 1068 (2021).
(2024). [36] C. L. Kane and E. J. Mele, Quantum Spin Hall Effect in
[22] Y. Zhang, T.-T. Tang, C. Girit, Z. Hao, M. C. Martin, Graphene, Physical Review Letters 95, 226801 (2005).
A. Zettl, M. F. Crommie, Y. R. Shen, and F. Wang, [37] S. Konschuh, M. Gmitra, D. Kochan, and J. Fabian, The-
 8

ory of spin-orbit coupling in bilayer graphene, Physical

Review B 85, 115423 (2012).
[38] A. Kurzmann, Y. Kleeorin, C. Tong, R. Garreis,
A. Knothe, M. Eich, C. Mittag, C. Gold, F. K. de Vries,
K. Watanabe, T. Taniguchi, V. Fal’ko, Y. Meir, T. Ihn,
and K. Ensslin, Kondo effect and spin–orbit coupling
in graphene quantum dots, Nature Communications 12,
6004 (2021).
[39] R. M. Potok, I. G. Rau, H. Shtrikman, Y. Oreg, and
D. Goldhaber-Gordon, Observation of the two-channel
Kondo effect, Nature 446, 167 (2007).
[40] F. D. M. Haldane, Nonlinear Field Theory of Large-Spin
Heisenberg Antiferromagnets: Semiclassically Quantized
Solitons of the One-Dimensional Easy-Axis N\’eel State,
Physical Review Letters 50, 1153 (1983).
[41] F. D. M. Haldane, Continuum dynamics of the 1-D
Heisenberg antiferromagnet: Identification with the O(3)
nonlinear sigma model, Physics Letters A 93, 464 (1983).
 9

Appendix A: Sample Design

(a) (b) RB
PG QD
LB
SGC1

SGM AlOx
top SGB
T M hBN
SG SG SGC2 bottom BLG
hBN
back
gate
B
SG

7.5 µm

QD
(c) (d) heater 1 (e) heater 2

Idet SGC1
Iheater
SGT SGM
T SGC1
SGM
FGC1
RB

2
Iheater

FGC
PG
200 nm LB 200 nm 200 nm
SGB SGC2

FIG. 5. (a) Optical microscope image of the used sample. The BLG flake is outlined in white. The applied back gate
voltage is negative, rendering the BLG p-type/hole conducting. The split gate pairs SGM, SGB and SGM, SGT are used to
electrostatically define a conducting channel. (b) Schematic of the VdW stack structure in the region where the QD is defined.
AFM images of the (c) QD (black circles) formed with finger gates LB, PG, RB and the charge detector tuned with finger
gate T, and the (d) (e) heater gate structures to form the ohmic constrictions, which are formed with split gates SGM, SGC1,
SGC2 and finger gates FGC1 and FGC2.
 10

(a)

(-1,1,0) (-1,0,0) (-1,0,-1)

(0
,0
,0 (
(0 ) 0,0
,1 ,-1
,0 )
)

(0,1
,-1)

(b)
(0,0,0) (0,1,0) (-1,0,0) (0,0,-1) (-1,1,0) (0,1,-1) (-1,0,-1)

FIG. 6. (a) Charge detector current as a function of left and right barrier voltage. The plunger gate voltage is fixed to 3.25 V.
Three types of resonances, corresponding to three types of QDs are observed: 1. pn-junction defined electron QD under LB
corresponding to the horizontal transconductance peaks. 2. barrier defined hole QD between LB and RB corresponding to the
bent transconductance peaks. This is the QD used for the entropy measurements. 3. pn-junction defined electron QD under
RB corresponding to the vertical transconductance peaks. The number triplet labels the occupation number of the respective
QDS (1., 2. 3.). (b) Diagrams showing the band bending induced by the finger gates LB, PG, RB in order to form the different
types of QDs for a few exemplary cases. QDs form where the conduction band edge Ec or valence band edge Ev crosses the
chemical potential µ in energy.
 11

Appendix B: Excited state spectra for one & two carriers

Here, we put forward a model that captures the effect of Kane-Mele type SOI, as present in BLG, on the energy
spectrum of the second charge carrier state. We assume that the carriers are in there orbital ground state. The valley
quantum numbers are denoted as K + (valley plus), K − (valley minus), and the spin quantum numbers as ↑ (spin
up), ↓ (spin down). We focus on the Hilbert space spanned by valley- and spin degree of freedom, i.e. H = Hv ⊗ Hs .
One can form valley-singlet and -triplet states:
1
|Sv ⟩ = √ (|K + K − ⟩ − |K − K + ⟩) (B1)
2
|Tv ⟩ = |K − K − ⟩
−
(B2)
1
|Tv0 ⟩ = √ (|K + K − ⟩ + |K − K + ⟩) (B3)
2
|Tv+ ⟩ = |K + K + ⟩ (B4)
as well as spin-singlet and -triplet states:
1
|Ss ⟩ = √ (|↑↓⟩ − |↓↑⟩) (B5)
2
|Ts ⟩ = |↓↓⟩
−
(B6)
1
|Ts0 ⟩ = √ (|↑↓⟩ + |↓↑⟩) (B7)
2
|Ts+ ⟩ = |↑↑⟩ (B8)
The Hamiltonians operating in H describing the one- and two-particle energy spectra are then given by
|K − ↓⟩ |K + ↑⟩ |K − ↑⟩ |K + ↓⟩
(1)
− 12 (gv + gs )µB B⊥ 0 0 0
 
1 (1)
0 2 (gv + gs )µB B⊥ 0 0
 
(1)
HES =  (B9)
 
(1)
0 0 ∆SO − 21 (gv − gs )µB B⊥ 0

 
(1)
0 0 0 ∆SO + 21 (gv − gs )µB B⊥

|S T 0 ⟩ |Sv Ts− ⟩ |Sv Ts+ ⟩ |Ss Tv− ⟩ |Ss Tv+ ⟩ |Ss Tv0 ⟩
 v s
0 0 0 0 0 ∆SO

 0 −gs µB B⊥ 0 0 0 0 
 0 0 gs µB B⊥ 0 0 0 
 
(2)
HES =
 0 (2) (B10)
0 0 0 0 

J1 − gv µB B⊥
(2)
 
 0 0 0 0 J1 + gv µB B⊥ 0 
∆SO 0 0 0 0 J1 + J2
Diagonalization leads to the eigen states and -energies compiled in Tab. I and Tab. II. The spectrum is visualized
in Fig. 7. The Kane-Mele SOI induced splitting between the new ground state and the first excited state of the
two-carrier charge state is given by
s 2
J1 + J2 J1 + J2
∆′SO = + ∆2SO − >0 (B11)
2 2
The angle θ characteristic for the superposition that forms the new ground state is given by the relation
2∆SO
tan(θ) = (B12)
J1 + J2
From this one can see that the strength of mixing between states |Sv Ts0 ⟩ and |Tv0 Ss ⟩ depends on how large the Kane-
Mele spin–orbit gap ∆SO is compared to the exchange splittings J1 , J2 . For ∆SO ≪ J1 + J2 one obtains |Sv Ts0 ⟩ as
the ground state, whereas for ∆SO ≫ J1 + J2 the ground state is given by the superposition √12 (|Sv Ts0 ⟩ − |Tv0 Ss ⟩).
Applying an out-of-plane magnetic field B⊥ does not cause a change in ground state for the N = 0 and N = 1
charge carrier states. For the N = 2 charge carrier state there is a ground state crossing happening at field B × . For
B⊥ > B × the new ground state is given by |Tv− Ss ⟩. Further increasing the magnetic field leads to higher orbital
states coming down in energy. This is not shown in Fig. 7.
 12

TABLE II. Excited state spectrum of the two-carrier charge

state.
TABLE I. Excited state spectrum of the one-carrier charge eigen states
(2)
eigen energies (w.r.t. E0 )
state. cos( θ2 ) |Sv Ts0 ⟩ − sin( θ2 ) |Tv0 Ss ⟩
eigen states
(1)
eigen energies (w.r.t. E0 ) =: |Sv Ts0 ⟩ ⊕ |Tv0 Ss ⟩ 0
(1) |Sv Ts− ⟩ ∆′SO − gs µB B⊥
|K − ↓⟩ − 12 (gv + gs )µB B⊥
(1) |Sv Ts0 ⟩ ∆′SO
|K + ↑⟩ + 2 (gv + gs )µB B⊥
1
|Sv Ts+ ⟩ ∆SO + gs µB B⊥
′
(2)
(1) |Ss Tv− ⟩ ∆′SO + J1 − gv µB B⊥
|K − ↑⟩ ∆SO − 12 (gv − gs )µB B⊥ (2)
(1) |Ss Tv+ ⟩ ∆SO + J1 + gv µB B⊥
′
|K + ↓⟩ ∆SO + 21 (gv − gs )µB B⊥ |Ss Tv0 ⟩ ∆′SO + J1 + J2
sin( 2 ) |Sv Ts ⟩ + cos( θ2 ) |Tv0 Ss ⟩
θ 0

=: |Sv Ts0 ⟩ ⊞ |Tv0 Ss ⟩ 2∆′SO + J1 + J2

(a) (b)

FIG. 7. Energy spectrum without (a) and with (b) considering the effect of Kane-Mele type SOI for the two-carrier charge
state N = 2.
 13

Appendix C: Finite bias spectroscopy

(a) (b)

FIG. 8. (a) Coulomb diamond measurement at the 1 → 2-transition. The transconductance peaks show the ground state
and an excited state separated in energy ∆SO from the ground state. The lever arm for the first carrier is extracted from this
measurement. (b) Finite bias measurement in out-of-plane magnetic field. The ground state splits into two states, the Kramers
pair |K − ↓⟩, |K + ↑⟩. The excited state splits into the Kramers pair |K − ↑⟩, |K + ↓⟩. From the slopes we extract the g-factors
(1) (1)
for the valley- gv and spin-Zeeman effect gs for the first carrier. All extracted parameters are compiled in Tab. III.

TABLE IV. Parameters from 1 → 2 transition

TABLE III. Parameters from 0 → 1 transition (2)
leverarm αPG = 0.025
lever arm
(1)
αPG = 0.025 SO induced splitting ∆′SO = 45 µeV
SO splitting ∆SO = 73 µeV exchange splitting J1 = 160 µeV
(1)
valley g-factor
(1)
gv = 13.6 valley g-factors first carrier gv = 15.5
(2)
spin g-factor
(1)
gs = 2.0 valley g-factor second carrier gv = 16.0
(2)
spin g-factor gs = 2.0
 14

(a) (b)

FIG. 9. (a) Coulomb diamond measurement at the 1 → 2-transition. The transconductance peaks show the ground state and
two excited states: one for positive bias separated by theenergy ∆′SO from the ground state, one for negative bias separated by
the energy ∆′SO + J1 from the ground state. The lever arm for the first carrier is extracted from this measurement. (b) Finite
bias measurement in out-of-plane magnetic field. The ground state does not split. The excited state with energy separation
of ∆′SO only visible for positive bias moves down in energy with a g-factor of 2.0, indicating that it is the state|Sv Ts− ⟩. The
excited state with an energy separation of ∆′SO + J1 only visible for negative bias moves down in energy with a g-factor of 16.0,
indicating that it is the state|Sv Ts− ⟩. All extracted parameters are compiled in Tab. IV.
 15

Appendix D: Thermodynamic Model

1. Grand canonical ensemble

QD is coupled to a reservoir, allowing for carrier and energy to fluctuate. If the coupling is dominated by thermal
broadening, i.e. kB T ≫ ΓR , the thermodynamics of the system is in general described in the framework of the grand
canonical ensemble. The grand partition function is given by
∞ X
∞ E
(N )
−N µ
X − i
Z(µ, T ) = e kB T
(D1)
N =0 i=0
∞ E
(N ) ∞
−N µ X ∆
(N )
− 0k T − ki T
X
= e B e B (D2)
N =0 i=0

where we introduced the decomposition

(N ) (N ) (N )
Ei = E0 + ∆i (D3)

(N ) (N )
where Ei is the energy of the ith excited state of the N -charge carrier state, E0 the ground state energy of the
(N )
N -charge carrier state, and ∆i the energy difference between the respective excited state and the ground state. All
macroscopic state quantities such as N (µ, T ) and S(µ, T ) can be calculated from Z(µ, T ), as well as the occupation
(N )
probability pi (µ, T ) of the microstate labeled (N, i). In terms of Z(µ, T ) one finds

(N ) ∂ ln Z(µ, T )
pi (µ, T ) = −kB T (N )
(D4)
∂Ei
∂ ln Z(µ, T )
N (µ, T ) = kB T (D5)
∂µ
∂ ln Z(µ, T )
S(µ, T ) = kB ln Z(µ, T ) + kB T . (D6)
∂T

and in terms of microstates/microstate probabilities one has

(N )
E −N µ
− 0 (N )
∆
(N ) e kB T
− i
pi (µ, T ) = e kB T
(D7)
Z(µ, T )
∞ ∞
(N )
X X
N (µ, T ) = N pi (µ, T ) (D8)
N =0 i=0
∞ X ∞
(N ) (N )
X
S(µ, T ) = −kB pi ln(pi ) (D9)
N =0 i=0

2. Entropy at zero magnetic field

(N −1)
For the charge transition N − 1 ↔ N contributions of other charge states can be neglected if kB T ≪ E0 −
(N −2) (N +1) (N )
E0 , E0 − E0 . The grand partition function becomes

E
(N −1)
−(N −1)µ ∞ ∆
(N −1)
E
(N )
−N µ ∞ ∆
(N )
− 0 X − i − 0 X − i
Z (N −1↔N ) (µ, T ) = e kB T
e kB T
+e kB T
e kB T
(D10)
i=0 i=0

(N ) (N −1)
If additionally kB T ≪ ∆i , ∆i one can also neglect excited states of the respective charge state in the expression
for the grand partition function. Under these considerations, the grand partition function, the microstate probabilities,
and the entropy are for the 0 → 1-transition given by
 16

(0) (1)
E E −µ
− k0T − 0
Z (0↔1)
=e B +e kB T
·2 (D11)
(0) 1
p0 = (1)
(D12)
µ −µ
− 0
kB T +ln(2)
1+e
(1) (1) 1 1
p|K − ↓⟩ = p|K + ↑⟩ = (D13)
2 µ
(1)
0
−µ
kB T −ln(2)
1+e
h i
(0) (0) (1) (1) (1) (1)
S (0↔1)
= −kB p0 ln(p0 ) + p|K − ↓⟩ ln(p|K − ↓⟩ ) + p|K + ↑⟩ ln(p|K + ↑⟩ ) (D14)

and for the 1 → 2-transition

(1) (2)
E −µ E −2µ
− 0 − 0
Z (1↔2) = e kB T
·2+e kB T
(D15)
(1) (1) 1 1
p|K − ↓⟩ = p|K + ↑⟩ = (D16)
2 −
µ
(2)
0
−µ
kB T −ln(2)
1+e
(2) 1
p|Sv T 0 ⟩⊕|T 0 Ss ⟩ = (2)
(D17)
s v µ −µ
0
kB T +ln(2)
1+e
h i
(1) (1) (1) (1) (2) (2)
S (1↔2)
= −kB p|K − ↓⟩ ln(p|K − ↓⟩ ) + p|K + ↑⟩ ln(p|K + ↑⟩ ) + p|Sv T 0 ⟩⊕|T 0 Ss ⟩ ln(p|Sv T 0 ⟩⊕|T 0 Ss ⟩ ) (D18)
s v s v

Note that here the kets merely serve as labels. Only the ground state degeneracies of the respective charge states
and the temperature determine the line shape of these functions.

3. Entropy at fixed particle number N

By putting the chemical potential µ far away from any transition, the particle number N is fixed. The QD is in
the N carrier state with absolute certainty. A description in the canonical ensemble framework is appropriate. The
partition function is then given by

∞ (N )
∆
j
−
X
Z (N )
(T ) = e kB T
(D19)
j=0

The probabilities of the microstates being realized is

(N )
∆
− i
(N ) e kB T
pi (T ) = (N )
(D20)
∞ ∆
j
P −
e kB T

j=0

4. Microstate probabilities of the BLG QD

For the N = 0 carrier state we find

Z (0) = 1 (D21)
(0)
p0 =1 (D22)
S0 = 0 (D23)
 17

With Tab. I we find for the N = 1 carrier state

(1) (1)
(gv +gs )µB B⊥ (gv +gs )µB B⊥
Z (1) = e 2kB T
+e 2kB T
(D24)
(1)
(gv +gs )µB B⊥

(1) e 2kB T
p|K − ↓⟩ = (D25)
Z (1)
(1)
(gv +gs )µB B⊥
−
(1) e 2kB T
p|K + ↑⟩ = (1)
(D26)
hZ i
(1) (1) (1) (1)
S1 = −kB p|K − ↓⟩ ln(p|K − ↓⟩ ) + p|K + ↑⟩ ln(p|K + ↑⟩ ) (D27)

where we included only the states lowest in energy as ∆SO > 4kB T in the experiments. For the N = 2 carrier state
we find with Tab. II:
(2)
∆′ +J1 −gv µB B⊥
− SO
Z (2)
=1+e kB T
(D28)
(2) 1
p|Sv T 0 ⟩⊕|T 0 Ss ⟩ = (2) (D29)
s v Z
(1)
∆′ +J1 −gv µB B⊥
− SO
(2) e kB T
p|T − S = (2)
(D30)
s⟩
h Z
v
i
(2) (2) (2) (2)
S2 = −kB p|Sv T 0 ⟩⊕|T 0 Ss ⟩ ln(p|Sv T 0 ⟩⊕|T 0 Ss ⟩ ) + p|T − S ⟩ ln(p|T − S ⟩ ) (D31)
s v s v v s v s

where we included only the states lowest in energy as ∆′SO > 4kB T and states causing a ground state crossing at finite
B⊥ . With these expressions we define

∆S0→1
calc
= S1 − S0 (D32)
∆S1→2
calc
= S2 − S1 (D33)