The first cuprate discovered by Bednorz and Muller has been the family of 124 compounds [3]. The parent La₂CuO₄ is an AF Mott insulator with T _N = 240 K that can be doped by substitution of a concentration x of La³⁺ by Ba²⁺ or Sr²⁺ [4]. The long-range ordered AF state is destroyed for x = 2%, and an SC dome appears for 0.07 < x < 0.3 with a maximum T _c = 38 K for x ∼ 0.19. From simple chemical arguments, the concentration of holes in the CuO₂ planes is merely n _h = x and most researchers have immediately considered that this (T,n _h ) phase diagram is generic of the cuprate plane. One noticed as well that the material displays in the intermediate range 0.02 < x < 0.07 a low T disordered frozen spin state of the Cu 3d ⁹ magnetic sites labeled as spin glass (SG).

A year later, the YBa₂Cu₃O_6+x family has been discovered [5], with a parent AF state as well with T _N = 410 K for x = 0 and part of a SC dome with a maximum T _c = 93 K. This compound involves two CuO₂ layers with an intermediate Y³⁺ layer. Successive bilayers are separated by a square lattice layer of non-magnetic Cu⁺ for x = 0. The hole doping for x > 0 results from the extra oxygen atoms introduced in this intermediate Cu layer. The relation between x and n _h is not straightforward as the Cu²⁺-O^2-- Cu²⁺ chain segments formed do not transfer charge in the CuO₂ planes. The phase diagram displayed in Figure 1 is quite similar to that of La₂CuO₄, but the AF phase is contiguous to the SC phase [7]. An important difference is therefore the absence of an intermediate SG regime. Although this is a very important distinction, many researchers mostly interested by the SC properties considered that as a minor difference and mapped the n _h ranges in the 124 and YBCO phase diagrams by an artificial parabolic fit of the SC domes [8].

In our preliminary experiments with Zn impurities in YBa₂Cu₃O_6+x , we immediately found that the reduction of T _c was larger in the underdoped regime than in the optimally doped case. This Zn induced in plane disorder has then led us to notice changes in the morphology of the phase diagram (PD), as displayed in Figure 1 [9]. Then, we plotted together the PD versus oxygen content x for pure samples and for 4% Zn substitution [6]. One can see that both extents of the AF phase and the SC dome shrink and disclose a wide SG regime. The morphology of the PD resembles then markedly that of the 124 family upon this introduction of disorder. The latter, therefore, plays a major role in defining the extent of the SG regime in the cuprates. This observation also implied that the n _h mapping between phase diagrams is not accurate.

This occurred at the very time of the PG discovery in the underdoped pure YBCO using ⁸⁹Y NMR shift data [10]. I initiated these Zn substitution experiments mostly to try to understand whether the PG had any relation with superconductivity. With 4% Zn substitution, our ⁸⁹Y NMR data immediately convinced us that T* unchanged even for the underdoped YBCO_6.6 sample, for which SC is fully suppressed [9]. This total independence of the two effects suggested that the PG is a phenomenon that competes with the SC state. This powerful indication revealed by introduction of Zn impurities does not, however, allow to conclude whether T* corresponds to a phase transition in the defect-free compound. The actual disorder introduced by the Zn substitution might indeed prevent a long-range ordering.

We were, at that time, most concerned by an important question concerning the PG. What could explain the large T _c difference at optimal doping between the 124 and YBCO families? Could that be associated with an intrinsic difference between compounds with single CuO₂ layers and those with bilayers in the unit cell? We completely cleared that question by performing the first study of powder samples of the single-layer Hg1201 cuprate family, for which the optimum T _c is approximately 85 K. We could synthesize samples covering a large range of doping from the underdoped to the overdoped regime. The comparison of ¹⁷O NMR shift data with that taken in the YBCO family allowed us to demonstrate that PG T* are, in all aspects, quantitatively identical in both families [11].

In this Hg cuprate family, doping proceeds again from an increase in the O content in the Hg layer that is far from the CuO₂ plane so that the dopant-induced disorder should be weaker than in the 124 family. These experiments confirmed, therefore, that the PG is a robust phenomenon independent not only on the number of CuO₂ layers but also on the disorder induced by dopants or substitutions.

The Zn substitution experiments of Figure 1 indicate that the SC dome shrinks with increasing plane disorder. This was shown as well using electron irradiation to increase progressively the in-plane defect content (see Section 3). These experiments have led us altogether to understand that the morphology of the PD evolves with the increase in disorder induced by the increase in the Zn content. We could then propose the 3D phase diagram of Figure 2, in which the third axis represents the increase in the Zn content [12]. Transport experiments given in detail in Section 3 allowed to follow the iso-n _h red lines plotted in Figure 2 that confirm the average displacement of the SC dome to the right in the PD while the SG range broadens and the PG T* is unchanged.

We shall consider hereafter the various cuprate families and try to anticipate where their phase diagram could be located on such a 3D scheme. We discarded there the PD for homovalent substitutions of rare earths on the Y³⁺ sites in YBCO_6+x as they do not significantly modify T _c and the PD. We did not consider either the peculiar case of Pr that was shown to display a 4+ valence that changes the oxygen content and hole doping for both AF and SC [13, 14]. For heterovalent substitutions, one can assume that the disorder due to the Coulomb interaction with the dopants decreases as their distance to the CuO₂ planes increases.

For all the single-layer or bilayer cuprate families, the PG T* variation observed by various experimental techniques were found to match with those observed for YBCO. For those families that display T _c  ∼ 90 K at optimal doping, the dopants are located in the interlayers. So for these families, the mapping of the (T _c , n _h ) SC domes usually assumed by parabolic fits might be acceptable. In Figure 2, their SC domes are located very near that of YBCO.

Let me point out that a weak dopant-induced disorder remains in those actual compounds, for which the optimum T _c value is approximately 90 K. The AF and SG domains for those families could not be probed except for YBCO as they could not be sufficiently underdoped experimentally. We cannot say, at this stage, whether in the absence of disorder, the T _c dome would reach T _c = 0 before the onset of the AF state. Would the clean phase diagram display a first-order transition from the AF to the SC state? In Figure 2, we introduced, therefore, the question mark near the 2D phase diagram for vanishing disorder.

The heterovalent substitutions in (LaSr)124 and (La-Bi)2201 families occur on sites adjacent to the CuO₂ planes. We assumed that the disorder is large in these systems, and we consequently located their 2D PD along the “disorder” axis at points for which their SG ranges approximately match those for YBCO + Zn.

We also considered another PD for Y_1-y Ca_yBCO₆, in which the n _h doping proceeds from a partial substitution of Ca²⁺ on the Y³⁺ site inserted between the CuO₂ layers in the parent YBCO₆ [15]. We again found that the AF state disappears for a Ca content y = 2n _h of about 7%. A SG range was disclosed for 0.03< n _h < 0.09, and a maximum T _c value of approximately 30 K is reached for y = 0.25 that unfortunately appears to be the Ca chemical solubility limit. Here, the large disorder induced by the substitution site between the CuO₂ planes again generates a PD that maps even quantitatively that of (LaSr)124 for n _h < 0.15. We did not plot it in Figure 2 to keep the figure clear.

In our proposed 3D PD, the quantitative definition of the disorder for a given family is, of course, rather rough. The disorder magnitude depends on the defect types and even evolves with doping inside many cuprate families. The data for a given family should be on a distorted surface in Figure 2 rather than on parallel flat planes.

The 3D representation, however, gives a reliable indication on the actual trend followed in the real materials. The systematic analysis conducted allows us to conclude that the disorder governs the morphology of the (T, n _h ) phase diagram by opening the spin glass regime.

This 3D phase diagram even suggests that the disorder governs the optimal T _c value of the diverse cuprate families. In-plane disorder produced by Zn leads to a subsequent decrease in optimum T _c . The out-of-plane disorder also has an influence on T _c values, as illustrated for instance within the 124 family. Then, T _c  ∼ 45 K are obtained by electrochemical intercalation of excess oxygen [16] that induces less disorder than Sr²⁺ substitution on the La³⁺ sites.

From the above phase diagram analysis, we cannot, however, conclude whether the difference of optimal T _c is solely due to the existing disorder. Further data will allow us to discuss again that point later in Section 5.

Here, I did not discuss so far the interesting low T-phases detected well below T* that have been thoroughly studied during the last decades. Soon, after the cuprate discovery, a stripe charge order has been highlighted in the 124 phase [17]. It is strongly pinned for Ba²⁺ substitution on the La³⁺ site and fully depresses the SC state for x = 0.12. For Sr²⁺ substitution, this stripe phase is less pinned and only seen as a small dip in the T _c (n _h ) variation.

In YBCO, many unexpected phenomena were detected at low T in high-quality samples with x ∼ 0.6, starting by quantum oscillations on the resistivity [18], corresponding to a small Fermi surface orbit and a negative Hall effect [19]. Then, NMR in a high applied field allowed to show a charge ordering below T _c = 60 K [20]. The latter could possibly be associated with the 2D CDW order detected for T as high as 150 K by RIXS experiments in zero field [21]. Extension of these measurements in high field [22] provide final evidence that the 2D CDW transforms below 60 K into the 3D CDW seen by NMR. The present understanding is that a reconstruction of the Fermi surface occurs below the SC dome, in a narrower dome-shaped region of the phase diagram [19] centered at about x = 0.6. This original coexistence of SC and CDW order parameters has been underlined and has justified extensive theoretical studies [23, 24].

Independently of the originality of this situation, I anticipated early on that these electronic orders occurred well below T* and, therefore, could not be responsible for the pseudogap that is highly related to Mott physics [25]. Nowadays, an emerging consensus is that the cuprates can display diverse low T electronic orders in the PG regime [19]. These compete with SC and are somewhat distinct in the diverse cuprate families, as found in 124 and YBCO_6.6 families.

I have always suggested that the actual disorder or the order introduced by the hole dopants plays a great role in selecting the ground states [25]. The CDW dome in YBCO occurs for a hole doping nearby that for which a slight singularity occurs in the T _c (x) variation, as shown in Figure 1. Structural studies have also shown that in this range of oxygen contents in YBCO, a Cu-O chain ordering occurs in the interlayer Cu plane [26]. I am, however, not aware of any careful experimental effort to search whether some correlation holds between the oxygen structural order and the CDW. In any case, I still suggest that these competing charge orders are, therefore, not generic of the ideal CuO₂ plane conceived by theoreticians.

In the 2D cuprates, the largest perturbations are induced, of course, by defects located in the CuO₂ planes. The substitution of Cu²⁺ by Zn²⁺ performed from the early days of HTSC introduces a spin vacancy without any charge defect. In the AF state of YBCO₆, these spinless defects only suppress spins on the AF lattice. This spin dilution effect is sufficient to explain the reduction of T _N with the increasing Zn content [32].

For a given Zn content, the introduction of oxygen on the intermediate Cu plane leads to the same hole doping, as for the pure material. Apart from the appearance of the SG phase, we could immediately detect an increase in the ⁸⁹Y NMR linewidth that increases with decreasing T down to T _c . Such broadening can only be associated with the appearance of local magnetic moments [9].

In that regime when the oxygen content is large enough, the spin glass temperature T _SG becomes negligible above x = 0.6, as shown in Figure 1. Spinless impurities never induce such paramagnetic local states in standard independent electron metals. SQUID measurements of the spin susceptibility allowed a determination of an upper limit of its magnitude [33]. An increase in the NMR experimental sensitivity enabled us then to reveal for x = 0.6 weak ⁸⁹Y NMR signals shifted from the main ⁸⁹Y NMR line, with a T-dependent NMR shift. We could attribute those “satellite” lines to ⁸⁹Y nuclei located near the Zn substitution site [34].

The characteristic pseudogap T dependence of the NMR shift has been found unmodified for sites at some distance from the Zn impurities, up to 4% Zn content. This suggests that the establishment of a robust PG behavior requires at most a sample size of about 25 unit cells. The Zn-induced paramagnetic moments appear then as a cloud of Cu sites around Zn with at most a similar spatial extent. They are characteristic of the correlated electronic nature of the PG regime.

In the slightly overdoped YBCO₇ samples, a weak broadening of the Y NMR line has been detected but no satellite could be resolved in this strange metal regime above T* and T _c so that the magnitude and/or the spatial extent of the moments is weaker/smaller.

We then performed an important step in order to test the eventual influence of a charged defect. We used Li⁺ substitution on the Cu²⁺ site of the CuO₂ planes and surprisingly discovered that the ⁸⁹Y NMR spectra were identical for 1% substitution of Li or Zn [35]. Both types of defects induced the same paramagnetism in their vicinity. The extra electron given by Li, therefore just fills a hole state in the bath band. This only slightly modifies the overall hole content for dilute Li substitution. This experimental result allowed us to show that the dominant defect in the CuO₂ plane is the spin vacancy associated with the substituent rather than its local charge.

As this was remarkable, we performed a similar comparison in another correlated electron system, Y₂BaNiO₅, in which Ni²⁺ is ordered in chains of S = 1 spins that constitute a prototype of the Haldane chain [2]. Here, the substitutions on the Ni site disrupt the chains, a staggered paramagnetism occurs near the chain ends, and the ⁸⁹Y NMR then displays many satellite lines [36, 37]. These were identical for Zn, Cu, and Mg substitutions so that any disruption of the spin chain always induces the same staggered paramagnetism on the chain ends. The NMR data allowed a measure of the extension ξ _imp of this staggered perturbation. Its T dependence accurately fits the theoretically computed magnetic correlation length ξ for the pure chain.

Assuming a similar staggered local 2D paramagnetic cloud with an exponential correlation length for the cuprates, we have estimated that ξ _imp would increase from about two to five lattice constants in the PG range from T* to T _c for x = 0.6 [38]. The intrinsic width of the ⁸⁹Y NMR forbids any accurate fit with such a model for the x = 1 strange metal, but ξ _imp would not exceed two lattice constants and would be roughly T-independent. Such a model-dependent determination is rough but establishes a qualitative difference between the PG and strange metal regimes. These ξ _imp (T) would also agree with evaluations of the magnetic correlation length ξ determined from inelastic neutron scattering experiments in single crystals of pure YBCO [39, 40]. Therefore the staggered response detected around spinless impurities monitors the actual correlation length of pure cuprates, as found for the Haldane chain from both theory and experiments [2, 37].

We disclosed a further advantage of the Li substitution as the ⁷Li NMR signal magnitude and intrinsic resolution are larger than those for ⁶⁴Zn. We could, therefore, detect its NMR shift ⁷ K for the Li content as small as 1%, regardless of the hole doping in the metallic range of YBCO_6+x. This NMR shift ⁷ K results from the transferred hyperfine coupling from the four Cu nearest neighbors of Li in the CuO₂ plane. These are the Cu sites exhibiting the largest paramagnetic response, in an exponentially staggered paramagnetic cloud model. We could measure accurately the T variation in ⁷ K for any hole doping [35]. As shown in Figure 3A, it can be fitted quite accurately with a Curie Weiss dependence ⁷ K-K ₀ = C (T+Θ) ⁻¹ .

The effective moment given by C is found independent of hole doping within experimental accuracy. The Weiss temperature Θ shown in Figure 3B exhibits a large change from the PG to the strange metal regime. Although Θ is very small in the PG regime, it abruptly increases through the PG “critical” doping point.

We did notice that this T dependence of the local magnetic response is somewhat analogous to that known for Kondo magnetic impurities in classical Fermi liquid metals. The evolution with hole doping of the apparent Kondo temperature Θ is similar to that found when changing 3d impurities in pure copper metal, for which the Kondo temperature T _K changes from a few mK for Mn to 30 K for Fe [41].

Let me point out that all these data given in Figure 3A were taken above T _c and that for x = 0.6, they do not show any singularity at 150 K, the onset temperature of the 3D CDW detected by RIXS experiments [21]. The variation in Θ with hole doping of Figure 3B is obvious even from the data taken above 150 K so that the magnetic response is not influenced by this charge order, if present in our samples. This response to Li impurities is, therefore probing a specific property of the PG regime. Similar studies in other cuprate families have not been performed so far as an intrinsic disorder due to chemical doping complicates the experimental situation. More material research efforts on Li substitutions in favorable cases such as BiSCO2212 or Hg1201 would be required.

I believe that this paramagnetic response to spinless impurity substitutions characterizes an abrupt evolution of the metallic state from the PG to the strange metal regime in cuprates. It remains an important challenge for any theoretical interpretation of the correlated electronic properties of cuprate physics. ¹⁷O NMR experiments show that this paramagnetic behavior persists in the superconducting state of the PG phase, while the Kondo-like screening is quite reduced in the SC state of the strange metal [42].

I shall consider hereafter the influence of impurities and vacancies in the CuO₂ plane on the transport properties in YBCO_6+x single crystals.

We have been able to disclose the influence of dilute in-plane point defects in the three considered “clean” materials that allowed us to probe, respectively, the PG, strange metal, and Fermi liquid regimes. The Curie–Weiss magnetic perturbation induced by spinless defects and the low T saturation of the carriers scattering behave quite similarly with Kondo effect in classical Fermi liquids. The sharp increase in the Kondo-like temperature Θ through the pseudogap line is a striking generic feature of cuprates.

For large concentrations of defects, their mutual interactions might induce a MIC. We could enhance defect clustering and show the importance played by the defect morphology in driving the insulating state. The comparison with data taken on “pure” lower T _c compounds demonstrates that the dopant out-of-plane disorder induces similar upturns of the resistivity and an apparent MIC. This confirms our anticipations performed from simple comparisons of the phase diagrams in Section 2.

Can we reduce the disorder effects to study the metal insulator transition in the small doping part of the phase diagram? Within our study of in-plane defects, we have shown that this disorder is well-quantified by the residual resistivity ρ ₀, but we do not know accurately the action of out-of-plane defects. The present results prove that those play an overwhelming role for low doping. The organization of the dopants certainly drives the stabilization of the family-specific low T-competing phases in the PG regime as well.

Magnetic point defects usually induce pair breaking in classical superconductors. Here, YBCO₇ exhibits a strange metal behavior with nearly zero residual resistivity in its normal state. Electron irradiation allowed us a full suppression of superconductivity, although a metallic state persists. As displayed in Figure 7, we found that T _c scales linearly with the residual resistivity ρ ₀ that scales itself with n _d , the defect content [46]. Such a behavior disagrees qualitatively with the Abrikosov–Gorkov (AG) variation [53] expected for pair breaking that implies a sharp downturn of T _c (n _d ) before reaching T _c = 0.

We provided further evidence for the inadequacy of AG from the low magnitude of the ΔT _c /ρ ₀ slope and the observed variation in the width of the SC transition. The latter, being due to the inhomogeneous distribution of defects, remains narrow and does not broaden when approaching T _c = 0.

The quasi linear measured variation in T _c (ρ ₀ ) is a somewhat unexpected result. One would hardly obtain it as accurately by other means, such as impurity substitutions. It surprisingly applies also for the underdoped YBCO_6.6 samples, as shown also in Figure 7, provided one takes for ρ ₀ , the high T-value of the extra resistivity.

Emery and Kivelson (EK) have proposed [54] that a peculiarity of the cuprates is their weak superconducting carrier density. They suggested that in such a case, phase fluctuations of the superconducting order parameter could be the source of T _c reduction. The phase coherent SC transition should then occur when the resistivity reaches a critical value [55].

In [46], we provided a detailed discussion indicating that such a correlation between T _c and ρ(T _c ) might be compatible with the quasi-linear variation. We were led, however, to conclude that the experimental situation probably requires taking both pair breaking and phase fluctuations into account. The influence of phase fluctuations should be more important for large defect contents and in the underdoped PG regime, but the existence of a single slope of ΔT _c /ρ ₀ at all defect contents appears awkward. I shall show hereafter that more information is available from quantitative studies of the magnitude of the SCF.

Nernst effect [56], diamagnetism [57, 58], or paraconductivity [59, 60] measurements have been used to study the SCF above T _c . The main difficulty in such experiments resides in subtracting the normal state or spurious contributions from the data. Nernst effect data involve vortex motion contributions, while spurious paramagnetic phases can counterbalance the diamagnetic signal.

We have conducted systematic investigations of the evolution of the cuprate phase diagrams with increasing controlled disorder in the CuO₂ planes. Our initial Zn substitution experiments demonstrated that the disorder opens a large spin glass regime, as displayed in the 3D phase diagram of Figure 2. The controlled increase in the vacancy content in single crystals allowed us then to show that SCF persist above T _c up to a temperature T’ _c that decreases less than T _c with increasing disorder. The normal state residual resistivity ρ ₀ increases markedly altogether and low T upturns appear mostly in the underdoped regime for large disorder.

All these features had been observed in the “pure” singular low-T _c cuprate families such as LSCO [48] or (LaBi)2201 [50]. Large residual resistivity ρ ₀ was clearly apparent in the data for those cuprate families, for which the large fields applied to suppress T _c disclosed resistivity upturns even at optimum T _c [49]. Nernst effect data revealed a large range of SCF above T _c in the underdoped rPG regime as well [56]. In LSCO, THz conductivity experiments have also been performed to probe the SCF range [66]. The maximum range of 10–20 K is smaller than that found by Nernst effect all-over the SC dome. All these results confirm our initial guess that had led us to propose the 3D phase diagram of Figure 2 and to anticipate the presence of a large disorder that diminishes T _c in those specific cuprate families.

Comparing ρ ₀ of these low-T _c compounds with those induced by electron irradiation in our YBCO samples even suggests that the linear relation, as shown in Figure 7, between T _c and ρ ₀ might have an extended validity. Choosing then to use the ρ ₀ value at optimal doping to fix the position of the PD of the diverse cuprate families on the disorder axis in Figure 2 would not modify it significantly [67]. In the phase fluctuation scenario, a relationship between T _c and ρ(T _c ) is indeed expected. This ensemble of arguments therefore leads me to conclude that the dopant disorder is mostly at the origin of the reduced T _c values in the cuprate families with anomalously low optimal T _c .

Once we have considered that we should put aside the low-T _c cuprates, the question raises as to whether we can decide what family better represents clean cuprates. For YBCO, Bi2212, Hg1201, and Tl2201 that display the highest T _c  ∼ 90 K, we noticed that the extrapolated ρ ₀ is nearly zero at optimal doping [67]. These families are therefore nearly equivalent in that doping range. We could indeed initially establish the universality of the transition T* between the PG and the strange metal regimes in YBCO and Hg1201 [11].

Our phase diagram of Figure 2 provides evidence that determination of the hole contents n _h by mapping the SC “domes” between different cuprate families is certainly incorrect, except for the higher T _c families. Unfortunately, it appears difficult to underdope the Tl2201 family that is the only “clean” cuprate, for which the overdoped to non-SC transition is accessible. As indicated in Section 4, we have not been able so far to perform a full study of the variation in the thermodynamic properties of the SC state through that transition.

YBCO_6+x is similarly the only family on which the metal-to-insulator transition occurs without a spin glass intermediate regime. We have, however, seen here that for x = 0.6, native disorder effects already appear as T’ _c remains large with respect to T _c . Low T upturns of the resistivity occur for oxygen contents x < 0.4 [12], but we do not know how the insulator state sets in. We do not have any reliable control on the influence of the oxygen disorder during this crossover. One cannot anticipate whether a first-order SC insulator transition would occur in a disorder-free cuprate, hence the question mark we have introduced in the 3D phase diagram in Figure 2.

So although huge progress has been performed since the discovery of the cuprates, we only have, at hand, a limited number of cuprates that allow us to explore independently some characteristic clean compositions in the phase diagram of Figure 2. We were not able to discover so far a batch that would play for the cuprates a similar role to that played by silicon single-crystal wafers for semiconductors.

I like to point out that the present critical concern is even more pertinent if one considers the actual disorder that occurs on the sample surfaces. The latter are investigated using novel techniques restricted to easily cleavable single crystals. ARPES [68] and STM [69] experiments allow to thoroughly study the doping dependence of the k-space surface band structures of Bi2212 crystals. One can beautifully see there the dopant disorder in the spatially resolved STM spectra [69] that have been helpful to correlate electronic structure with the local order. This unavoidable dopant disorder probably limits the energy resolution in ARPES spectra. The bulk and surface dopant disorder are not necessarily identical so that refined comparisons between bulk thermodynamic properties and surface energy spectra might be misleading. Local STM and NMR experiments have been achieved on Zn substituted cuprates. NMR could give information mostly on the bulk YBCO normal state properties, while STM permitted to probe the SC local states on the Bi2212 surface [70].

The change in the normal state transport properties has initially revealed the difference between the PG, strange metal, and quasi Fermi liquid regimes in the “pure” cuprates. The introduction of point defects has allowed us to highlight those differences by providing clear evidence that the scattering of the carriers splits into two distinct contributions. The contribution observed in the pure compound is totally unmodified by the defects introduced. It adds to the specific scattering of the carriers by the dilute impurities. An unexpected difference with the situation encountered in classical metals is that in the latter case, the scatterings are associated with two different processes, an elastic scattering on the defects and an inelastic one mediated by phonon modes. In the cuprates, both processes deal with electronic states controlled by the correlated electronic properties of the pure compound.

This stems from the fact highlighted by the NMR experiments that the impurity substitutions or vacancies in the CuO₂ planes induce a modification in their vicinity, which has a significant spatial extension that varies with temperature and doping [38]. This is certainly unambiguous evidence that the cuprates are strongly correlated electron systems.

We gave evidence for low T upturns of the resistivity as well when introducing in-plane vacancies by electron irradiation. These correspond to scattering rates on the defects similar to those observed for Kondo magnetic impurities in classical metals. As recalled in Section 3, these upturns scale with defect content as long as their concentration is small enough to avoid overlaps between the clouds induced by the vacancies.

We have shown, both by NMR and transport, that a sharp modification of the defect state occurs when the hole doping of YBCO is switched from the PG to the strange metal regime. We assign that to a reduction of the paramagnetism similar to a large increase in the Kondo-like temperature. In the strange metal regime, the cloud size does not exceed two lattice constants and the resistivity upturns decrease, in agreement with an increased “Kondo” temperature. We assume that paramagnetism should disappear when reaching the pure Fermi liquid regime for large hole doping. This crossover still needs, however, to be carefully studied for instance on the Tl2201 cuprate family.

The data analysis of Section 4 reports an original study of the destruction of the 2D superconductivity by a controlled disorder. The GL or Lawrence–Doniach pair breaking approach explains rather well the 2D SCF up to 1.1 or 1.2 T _c in the pure or disordered YBCO near-optimal doping. This would apparently also hold for very clean underdoped samples.

In that PG regime, any disorder native or produced by electron irradiation, induces a large SCF range as compared to T _c that mostly originates from a loss of phase coherence [60]. [55] had proposed a major importance of phase fluctuations for SC with either a low superfluid density or a large resistivity at T _c . Our study indicates that the disorder is mainly responsible for the phase fluctuations in underdoped YBCO. I therefore suggest that the Kondo-like inelastic defect scattering of the hole carriers is the main player in the T _c reduction that occurs before the SC insulator transition.

One should notice finally that our study of the field dependence of the SCF gives a measure of H’ _c (0), the critical 2D magnetic field. Thermodynamic arguments allow us to connect it with Δ_SC, the SC gap magnitude. In clean YBCO, Δ_SC increases with doping up to the optimum T _c , and correspondingly, the coherence length ξ(0) diminishes. One would expect to reach a cutoff value of ξ(0) and then a decrease in the gap in the overdoped regime. One should still probe that experimentally in the Tl2201 family and altogether whether the transition at the endpoint of the SC dome is smeared out by disorder.