Two-component uniform spin susceptibility in superconducting HgBa CuO single 2 4+δ crystals determined with 63Cu and 199Hg NMR Ju¨rgen Haase1,∗ Damian Rybicki1, Charles P. Slichter2, Martin Greven3, Guichuan Yu3, Yuan Li4, and Xudong Zhao3,5 1 Faculty of Physics and Earth Science, University of Leipzig, Linne´straße 5, 04103 Leipzig, Germany 2 Department of Physics, University of Illinois at Urbana-Champaign, Urbana IL 61801, USA 3 School of Physics and Astronomy, University of Minnesota, Minneapolis, Minnesota 55455, USA 4Max-Planck-Institut fu¨r Festko¨rperforschung, Heisenbergstrasse 1, D-70569 Stuttgart, Germany and 2 5College of Chemistry, Jilin University, Changchun 130012, China 1 (Dated: January 23, 2012) 0 2 63Cuand199HgNMRshiftsforanoptimallydopedandunderdopedHgBa2CuO4+δ singlecrystal n arereported,andthetemperaturedependencedictatesatwo-componentdescriptionoftheuniform a spin susceptibility. The first component, associated with the pseudogap phenomenon in the NMR J shifts, decreases already atroom temperatureandcontinuestodrop asthetemperatureislowered, 0 without adrasticchangeat thetransition temperatureintothesuperconductingstate. Thesecond 2 componentistemperatureindependentabovethesuperconductingtransitiontemperatureandvan- ishes rapidly below it. It is a substantial part of the total T dependent susceptibility measured at ] both nuclei. n o c I. INTRODUCTION shift measurements that La Sr CuO does not 1.85 0.15 4 - show a single-component spin response6, and since r p high quality HgBa CuO single crystals had become One of the most challenging questions in condensed 2 4+δ u available7,8, we address here the important question matter theory is embodied in the physics of high- s whether the two-componentbehavior is a mere peculiar- . temperature superconducting cuprates and concerns at their transition from a Mott antiferromagnetic insulat- ity of La1.85Sr0.15CuO4 or a more universal property of m the cuprates. ingstateintoaconductorandsuperconductorbyremov- - ing or adding electrons, i.e., by doping them. Clearly, We have acquired a comprehensive set of NMR data d understanding the electronic spin or magnetic spin sus- foranunderdopedandoptimallydopedsample9,andwe n ceptibility χ of these materials is of great interest. In will show that both materials require a two-component o S particular, it is important to know whether the uniform description. Concurrent with the present work, some of c [ spin susceptibility can be understood in terms of a sin- us were involved with developing a new anvil cell NMR gle electronic spin component that changes with dop- probe10 forsensitive,high-pressureNMRexperimentsto 2 ing and temperature across the phase diagram. Nuclear enable,e.g.,theinvestigationofthe cupratesunderpres- v 6 magnetic resonance(NMR) measures χS locally through sures that had not be done before. Recent success was 1 the hyperfine coupling that causes, e.g., a relative shift reported with 17O NMR of YBa2Cu4O8 at pressures up 0 K =p ·χ of the nuclear resonance frequency, where to 63 kbar11. It was found, based in part on the results S,η η S 6 p is the anisotropic hyperfine coupling constant1 and η presentedhere,thatthespinsusceptibilityinthissystem η 0. denotes the direction of the magnetic field with respect also requires a two-component description. 1 to the crystal axes. With NMR as a bulk probe, such 1 measurements are reliable since they are insensitive to 1 dilute impurities or magnetic inclusions. II. EXPERIMENTAL : v If all nuclei, n, of a given material experi- i ence the same normalized, proportional shift change The two crystals7 used in this study were annealed to X ∆nKS,η/npη =∆χSinthetemperatureinterval∆T,this result in one optimally doped crystal with Tc = 97 K r a isevidenceforasingle-componentresponse. Ontheother (mass 30.3 mg) that we label X97 and one underdoped hand, if one finds experimentally that this shift change crystalwithT =74K(mass3.3mg)thatwelabelX748. c with temperature differs for different nuclei, this proves 63Cuand199HgNMRshiftsasafunctionoftemperature the failure of a single-component view. This important T were recorded in a magnetic field B = 11.75 T for 0 question was addressed with NMR measurements early twodifferentorientationsofthetetragonalsinglecrystals on. And it was concluded, rather quickly, with exper- withrespecttotheexternalmagneticfield: withthefield iments on YBa2Cu3O6.632 and later YBa2Cu4O83 that parallel(ckB0)andperpendicular(c⊥B0)tothecrystal a single-component description of the cuprates is appro- c-axis. NMR shifts were referenced using a fine metallic priate. However, pioneering measurements of the total copper powder. For 63Cu we used the published value of susceptibility indicated two-component behavior of dif- the 63Cu Knight shift of 3820 ppm.12 199Hg shifts were ferent cuprates4, and NMR relaxation studies did not referenced to (CH ) Hg using the referencing procedure 3 2 agree with a single-component model5. described by Harris13. Below T , in particularfor 199Hg, c Recently, some of us showed with extensive NMR the sensitivity loss due to radio frequency penetration 2 1.65 a) b) X97 X97,II X97, 0.8 1.60 X74 1.60 0.7 63 K [%]1||.55 X74,II X74, 00..56 [%]K 63 63 K [%]II11..5505 1.50 0.4 1.45 1.45 0.3 0.4 0.5 0.6 0.7 63 0 100 200 300 0 100 200 300 K [%] Temperature [K] Temperature [K] FIG. 2. 63K (T) versus 63K (T) with T as an implicit FIG.1. T dependenciesofthetotal magneticshifts63K (T) k ⊥ k,⊥ parameter, for both single crystals, X74 and X97. The two for both crystals, X74 (Tc = 74K ) and X97 (Tc = 97K); a) arrows indicate Tc for both crystals. The two straight lines ckB0, and b) c⊥B0. The quadrupole interaction was deter- are linear fits to the data over the entire range for X97 and mined and its effect on theshift removed for c⊥B0. aboveTc for X74. and slow nuclear relaxation limits data collection. Since thematerialsareinamixedsuperconductingstatebelow La2−xSrxCuO417). At higher T the T dependence must T , a macroscopic diamagnetic contribution K that is then be due to the T dependent spin susceptibility. c D not precisely known appears. Since KD is isotope inde- Let us assume momentarily that the shifts in Fig. 1 pendentforagivensampleorientation,itcanberemoved are caused by a single component spin susceptibility by taking shift differences, e.g., 63Kη−199Kη. χS(T). Then, all four curves in Fig. 1 are described by 63K (x ,T) = 63p · χ (x ,T) where η denotes S,η j η S j the orientation of the crystal with respect to the mag- III. RESULTS AND DISCUSSION netic field and x , x denote the doping. If we plot 1 2 for a given sample the shifts measured for one orien- tation against that measured for the other orientation Fig. 1 shows the T dependence of the total magnetic Cu shifts 63K (the error bars are within the size of the we expect straight lines with similar slopes. This can η be seen by substituting χ (x ,T), e.g., 63K (x ,T) = symbols)thatarethesumoforbital,K ,andspin,K , S j S,k j orb S 63p /63p ·63K (x ,T). SuchplotsareshowninFig.2 contributions. Notethatthiscuprateisknowntoshowa k ⊥ S,⊥ j T dependent Cu shift also for ckB 14–16, in contrastto, (the actual data are the same as those presented in 0 e.g., YBa Cu O and La Sr CuO . The term due Fig. 1). We observe indeed straight lines with similar 2 3 7−δ 2−x x 4 to the partly diamagnetic response in the mixed state, slopes at higher T (larger shifts), however, there is an K , that we expect below T , can be neglected for the offset between them. Note that the similar slopes prove D c large Cu shifts, as we show below when we discuss the thattheratioofthehyperfinecoefficients,andthusmost 199HgNMRdata. Notethattheanisotropiesoftheshifts likely the coefficients themselves, is independent on dop- (comparetheactualnumbersinbothpanelsinFig.1)are ing (the same conclusion is reached when comparing Cu due to the orbital (Van Vleck) susceptibility. We show and Hg data, see below). For X97 the straight line con- the total magnetic shifts (verified with field dependent tinuesthroughTcdowntothelowesttemperatures,while measurements) to keep the discussion transparent. for X74 the measured data points move away from the straightlinenearT tomeetthosefortheX97materialat In order to isolate the spin shift one usually assumes c the lowest T. That they meet at low T is expected from thatK islargelyT independent(asarethequadrupole orb Fig. 1, the orbital shifts are the same for both samples. interaction and the hyperfine coefficients), and that the Note that while we expect constant slopes, the constant T dependence ofthe shiftisgivenbyaT dependentspin high-T offset between the two lines, which disappears at shiftthatvanishesbelowT . Thismeansχ (T →0)→0 c S low T, cannot be explained. duetospinsingletpairing,andthusthelow-T shiftsmust be given by the orbital term. Clearly, the observation in Fig. 2 that the changes of The fact that the shifts for X97 and X74 at the low- the X74 shifts with T are not proportionalto each other est T and given orientation are similar for each ori- atall T (here for differentorientationsofthe fieldrather entation in Fig. 1 shows that the orbital shift term thandifferentnuclei) demands a second spinsusceptibil- does not depend significantly on doping (similar to, e.g., itywithadifferentcouplingtothenucleus. Thus,weare 3 led to write6 Kη(xj,T)=pη·χ1(xj,T)+qη·χ2(xj,T), (1) arb. units] a) X97 where χ2(T) is the second T dependent spin susceptibil- ensity [ ity and qη the corresponding hyperfine coefficient. For a Int particularcrystal(xj)wethenhaveforthedatainFig.2, 89.2 89.3 89.4 89.5 89.6 89.7 Frequency [MHz] pk pk -0.38 b) c) 0.14 K (T)= ·K (T)+ q − ·q ·χ (T). (2) k p ⊥ (cid:20) k p ⊥(cid:21) 2 ,|| ⊥ ⊥ %]-0.40 0.12K 199 Then, χ2(X97,T)mustbe rathersmallandχ2(X74,T > K[|| , [% Tc)=const.,butvanishesbelowaboutTc,andbothsets 199 ] of data in Fig. 2 meet at low T. -0.42 0.10 The need for two shift components is already obvious ,|| inFig.1wheretheshiftsforbothorientationsoftheX97 -0.44 0.08 , crystalaredrawntoscaleinpanelsa)andb). Then,since the shifts in each panel are the same at low T the shifts -0.46 0.06 for both orientations of the X74 crystal (lower curves in 0 100 200 3000 100 200 300 both panels) should look very similar on this scale, as T [K] T [K] well. This is clearly not the case and Fig. 1 tells us that theshifts cannotbeunderstoodwithasingle-component FIG. 3. 199Hg NMR of the Tc = 97K single crystal (X97); a), typical spectra for both directions of the field show two spin susceptibility. Instead, we need at least two com- distinct Hg sites; b) and c), T dependencies of the shifts for ponents with different T dependences. We can also con- thelines shown above for ckB0 and c⊥B0, respectively. cludefromthe datainFig.1thatbothcomponentshave to disappear at low T. Wewouldliketopointoutthatthetwosusceptibilities thatcouplewithp 6=q tothenucleus,cf. (1),maynot η η be the suceptibilities of the actual spin components. For example, a second spin component that couples to the nucleus through the same p as the first one could be %] 0 100 200 300 η , [ 0.220 present in the shift data for the X97 crystal in Fig. 1 K, a) wfrohmileFniogt. s2hotwhaintgχu2pisininFdiegp.e2n.denHtowofevTer,abwoevedoTcinffoerr - 1.4* 00..221126 the X74 crystal. We refer to a more detailed discussion K, further below. %] 0.208 , [-0.004 We now turn to 199Hg NMR. We show in Fig. 3 data K, b) 8* foorrienthtaetiXon97socfrtyhsetafile:lda(ap)a,iranodftthyepTicadlepspenecdternacifeosrobfoththe - 0.-0.008 ,-0.012 correspondingshifts (b,c). Weobservetwowellresolved K 199Hg lines and label them with α and β. Note that the 0 100 200 300 T dependence of the Hg shifts is not very different from Temperature [K] that of Cu, but the shifts are smaller, as one expects from transferred hyperfine coupling to the more distant %] -0.027 295K Hg atoms. Similar results were obtained for the X74 [ c) K 100K crystal (data not shown), which reveal that the α line is 199 -0.030 thesameasfortheX97crystalatthelowestT wherethe spin shift is small, whereas the β line has a smaller shift -0.033 80K and lower intensity. Therefore, the β Hg site must be due to Hg atoms that are affected by doping (for further 0.036 0.039 0.042 199 details we refer to9). Here we would like to focus on the K [%] X97crystalinordertotestthesinglevs. twocomponent description for this system. FIG. 4. X97 crystal: a) , b) T dependence of Since we have two sharp, well resolved Hg resonances 199Kβ,k − 1.4 · 199Kα,k and 199Kβ,⊥ − 0.8 · 199Kα,⊥, wecanplottheirshiftsagainsteachother,similartowhat respectively. c) ∆199K⊥ ≡ 199Kβ,⊥ − 199Kα,⊥ vs. we did for 63Cu, to inquire about the spin susceptibility. ∆199K ≡199K −199K with T as an implicit parame- k β,k α,k Note,however,thattheorbitalcontributionstotheshifts ter(see text). are very large compared to the T dependent shifts, see 4 bothpanelsinFig.3. Itisthereforemostreliabletoplot the shifts of both lines α and β for each orientation η of the field against each other, i.e., 199Kβ,η vs. 199Kα,η. 295K From such plots (see below) we find straight lines with 7)1.60 9 fixed slopes for T >T . Since K =0 for T >T we infer X c D c ( thatthereisonlyoneT dependentspinsusceptibilityfor %] 2.0563 199 1.55 K - K T >Tc. We find for the slopes rk =1.4 and r⊥ =0.8 for K [|| 100K ckB0 and c⊥B0, respectively. In order to demonstrate 63 2.00 this withhighergraphicalresolutionweplotinFig.4(a, 1.50 b), 199K (T)−r ·199K (T) vs. T. Wecanlookat 1.95 β,η η α,η 70K (2)(cid:2)to discuss Fig. 4 (a, b). Sinc(cid:3)e rη = 199pβ,η/199pα,η, 1.45 horizontallinesforT >T inFig.4(a,b)areequivalent, 20K 1.90 1.92 1.94 1.96 c cf. (2), to constantslopesr aboveT inthe correspond- 1.44 1.46 1.48 1.50 1.52 1.54 η c 63 ingshift-shiftplots. BelowTc weobservedeviationsfrom K|| [%] (X74) the horizontal lines for both orientations, a clear indica- tion of another shift contribution. However, we cannot FIG. 5. Main panel: 63K (T) vs. 63K (T). k,X97 k,X74 conclude from it with (2) on a second T dependent spin In order to remove a possible diamagnetic term KD componentsinceKDcouldcausesuchabehavior,aswell. the Hg shifts have been subtracted from the Cu In order to remove the uncertainty due to KD we now shifts in the inset: (cid:2)63Kk,X97(T)−199Kk,α,X97(T)(cid:3) vs. take the differences between the shifts of the two lines (cid:2)63Kk,X74(T)−199Kk,α,X74(T)(cid:3). for each orientation. With ∆199K ≡199K −199K η β,η α,η we have with (1), here adopted for the 199Hg NMR spin shift, crystal (X97, vertical axis) plotted against the Cu shift with ckB of the other crystal (X74, horizontal axis). ∆199K =∆199p ·χ (T)+∆199q ·χ (T), (3) 0 η η 1 η 2 Surprisingly, one finds a straight line above the X97 T . c This says that, coming from high T (large shifts), the where ∆199p ≡ 199p − 199p , etc. Analogous to η β,η α,η shifts in Fig. 1 for both crystals fall in such a way that (2) we can express one shift difference in terms of the leaves the ratio of the slopes constant. At the T of the other in order to remove one of the susceptibilities, and c X97 crystal its shift drops rapidly and below 70 K we we have (dropping the mass number of the isotope for find a behavior similar to the one observed at higher convenience) T. The fact that they arrive at similar shift values for the lowest T has been noted in Fig. 1 and is expected ∆p ∆p for a doping independent orbital shift. Since we know ⊥ ⊥ ∆K = ·∆K + ∆q − ·∆q ·χ . (4) ⊥ ∆p k (cid:20) ⊥ ∆p k(cid:21) 2 that χ2(T > Tc) = const. we conclude from Fig. 5 for k k T >T thatsince[∆K ]/[∆K ]≈1.4wemusthave c X97 X74 This is shown in Fig. 4 (c), however,there we plot the ∆χ1(X97)/∆χ1(X74) ≈ 1.4 for all T intervals. Thus, difference of the total experimental shifts between the β χ1(X97) = 1.4·χ1(X74)+c, where c is a constant that and α line for c⊥B as a function of the same difference must have vanished below 70K in Fig. 5. This says that 0 for ckB0. Note that T independent orbital terms cause χ1 consists of a component that is T dependent at all T a constantoffset onboth axes. For a single T dependent and changes proportionally with doping, plus a second spin component, i.e., χ2 = 0 in (4), a straight line with component with a behavior similar to χ2. a fixed slope is expected. This is clearly not the case. Before we continue with the discussion we would like We observe that the shift difference for ckB0 drops sig- to provethatthe diamagneticshift KD is indeednotim- nificantly already at high T as the T is lowered, while portantwhencomparingCushifts. We showinthe inset that for c⊥B is constant at high T, but drops only be- ofFig.5aplotsimilartothatinthemainpanel,butwith 0 low about T , a behavior that cannot be explained by theHgshiftssubtractedfromthecorrespondingCushifts c a single T dependent spin susceptibility. From (4) we so that KD is eliminated on both axes. As expected, we conclude that we need a second susceptibility that is T observe again a straight line above Tc and since the Hg independent above T , but changes below T . This inde- shifts are small it has a similar slope. The drop just be- c c pendent experimental observation proves that a second low Tc remains and this rules out the importance of KD component is necessary also for the X97 crystal even if on the Cu shift scale. it appears to be small. While this is in agreement with We plot in Fig. 6 the Hg shifts (total experimental what we derived from comparing the Cu shifts for X97 shifts)oftheαlinesagainstthoseofCuforbothcrystals and X74, it was not obvious from the Cu shifts for the and c k B . As expected, the slopes at higher T are 0 X97 crystal alone. approximatelystraightlinesandsimilar(notethatlarger errors are expected since the Cu shifts are much bigger Somewhat unusual plots are presented in Fig. 5. In than those of Hg). We find that 199p /63p ≈0.12. k,α k the main panel we see the Cu shift with ckB for one For the X97 crystal we observe a steeper drop near T , 0 c 5 approximately also below T . We now understand why c the influence of the second spin component is rather -0.42 weak in the plots in Figs. 2 and 4 (c) for X97. Note, that we discovered with Fig. 5 that -0.43 ∆KX97/∆KX74 ≈1.4 at higher T. A simple explanation %] follows by assuming that χAA,X97 ≈ 1.4· χAA,X74 and [ that the integration constant is just due to χ . We AB,X K -0.44 then have with (5) for the curve in the main panel, 9 9 1 63K =r 63K -0.45 η,X97 η η,X74 X97 +63p ·χ X74 η AB,X97 (7) +63q ·[χ +χ ] -0.46 η AB,X97 BB,X97 1.45 1.50 1.55 1.60 63K [%] −rη 63pη+63qη χAB,X74+63qη·χBB,X74 . (cid:8)(cid:0) (cid:1) (cid:9) We note that the terms in the last row stem from the FIG. 6. X74 and X97 crystals: 199K vs. 63K for both X74 cyrstaland will not contribute to the drop in Fig. 5 k,α k crystals with T as an implicit parameter (see text). nearTc ≈97K.Inaddition,sincetheX74shiftforckB0 in Fig.1 does not changesignificantly below T we know c that the term in the last row can be ignored at lower T, which levels off to the slope above T before it finally as well. Similarly, the term in the third row was found c drops much more rapidly with lowering T. Since the to be rather small for the X97 crystal (χ2,X97 ≈ 0). We first drop below Tc for the X97 crystal is accompanied are left with the term in the second row, and since χAB by a large shift change for Cu, it is not dominated by vanishesrapidlyatTc ≈97Kthe observeddropinFig.5 K . The second and steep drop for both crystals at low is expected. We conclude that this drop is mainly given D T must be mostly due to KD since its slope is close to by 63pk·∆χAB,X97 ≈+0.05%. 1 (K is independent of the nucleus). With this we can We can also inspect Fig. 6 again. For the X97 crystal D estimate the maximum diamagnetic shift for ckB0 to the change of the slope at Tc marks the change in χAB be ∆KD < 0.015%. Thus, the contributions to the T andχBB untilKD dominatesatlowerT. Onthehorizon- dependent shift below Tc from KD and KS are ofsimilar tal axes this is given by ∆63Kk,X97 =63pk·∆χAB,X97 ≈ size for 199Hg (but not for 63Cu). +0.05%, as concluded from Fig. 5. The corresponding changeinshift forHg ofabout+0.01%is somewhatbig- Since we need two spin component (we call them A ger due to KD (199pk/63pk ≈ 0.12). For the X74 crystal and B) for the explanation of our data, we would like we do not expect a drop near Tc for Cu (cf. discussion to point out that this demands in general18 three spin above),and forHg it is mostly givenby KD. (It appears susceptibilities6: χAA,χBB, and χAB since there can be that the onset of the change in slope is slightly above Tc acouplingofthetwospincomponentsAandB.Accord- outside the estimated error in the T measurement). ingly, we have χ1 =χAA+χAB, and χ2 = χAB +χBB. On general grounds we expect χAA to be positive and The total spin susceptibility is then given by χS = therefore it follows that 63pη > 0 for both orientations. χAA+2χAB+χBB. For the NMR spin shifts it follows, Then, we have for T > Tc that χAB,X97 > 0, and since χ ≈−χ , it follow that χ < 0 above AB,X97 BB,X97 BB,X97 Kη,X =pη(χAA,X +χAB,X)+qη(χBB,X +χAB,X). (5) Tc for this material. For the X74 cyrstal we notice, e.g., from Fig. 1, that the constant shift term above T must c We know that (χBB,X +χAB,X) must be T independent be small for ckB0, but larger for c⊥B0. athigherT. Wenowassumethatbothcomponentχ BB,X and χ indeed contribute to the shifts. Since it is To conclude, based on a thorough NMR study of the AB,X highly unlikely that both components are T dependent, two HgBa2CuO4+δ crystals9,19 we identified a 2nd shift but their sum is not, we conclude that both must be T component that is independent of T above Tc and van- independent above Tc. Then, χAA must be the suscepti- ishes rapidly below Tc. Since it influences the Cu and bility that is T dependent already above T . Hg nuclei by similar relative values, the 2nd component c We deduce from Fig. 2 that p /p ≈0.40 and with mustbe due toaspinsusceptibility, inaccordwithwhat k ⊥ some of us proposed earlier for La Sr CuO 6, and 1.85 0.15 4 CX ≡ qk−(pk/p⊥)·q⊥ [χAB,X +χBB,X] (6) very recently for YBa2Cu4O811. Our results can be ex- (cid:2) (cid:3) plained consistently with two spin components that are that C ≈ 0%, C (T >T ) ≈ −0.040%, coupled to each other and thus have the three suscepti- X97 X74 c and C (T ≪T )≈ 0%. Thus, at high T bilities χ ,χ ,χ . In such a scenario, cf. (5), we X74 c AA AB BB [χ +χ ] ≈ const., but it vanishes rapidly find that only one of them, χ is T dependent above AB,X74 BB,X74 AA below T . We also find χ ≈ −χ , and if T ,whiletheothertwovanishbelowT . Furthermore,we c AB,X97 BB,X97 c c one of the components is sizable this relation holds findthatχ =1.4·χ ,arelationthathasbeen AA,X97 AA,X74 6 suggestedbysomeofusearlier19 andthatsomeofusap- review20 have put forward a model. Although the T de- plied successfully to understand the changes in the 17O pendence of the first component reminds one of coupled NMR spin shift of YBa Cu O with pressure11. Thus, Heisenbergspins4, wenote that the pseudogapregimeof 2 4 8 χ (T)isresponsibleforthepseudogapfeatureobserved the cuprates has recently been shown to be associated AA in NMR and it changes proportionally (increases) with with unconventional magnetism21,22. Whereas the 2nd increasing doping for an underdoped cuprate, but de- component exhibits the behavior of a Fermi liquid that creases with increasing pressure11. (We believe that becomes superconducting at T , a T constant behavior c χ (x )/χ (x ) = x /x where x is the local hole at higher T is also expected for a Fermi glass23. AA 2 AA 1 2 1 j doping per CuO in the plane, so that χ = ρ(T)·x, 2 AA see19). The part of the shift that is constant above T c was found to increase dramatically with pressure11 and IV. ACKNOWLEDGEMENT we believe that it mus also increase with doping. How- ever, more experiments are necessary to explore the two components across the phase diagram and as a function We acknowledge the help of M. Lux, M. Jurkutat, T. of pressure. Nevertheless, there can be no doubt that a Meissner, M. Bertmer, A. Po¨ppl; the financial support two-component scenario is generic to the cuprates. This by the University of Leipzig, and discussions with O.P. also means that the hitherto adopted single-component Sushkov (JH, DR). The crystal growth work was sup- interpretation of NMR data can no longer be upheld. portedbytheDepartmentofEnergyunderContractNo. We can only guess the physical origin of the two DE-AC02-76SF00515and by the NationalScience Foun- spin components. Barzykin and Pines in an extensive dation under Grant No. DMR-0705086. ∗ [email protected] 12 O. Lutz, H. Oehler and P. Kroneck, Z. Phys. A 288, 17 1 W. D. Knight,Phys. Rev.76, 1259 (1949) (1978) 2 M. Takigawa et al., Phys. Rev.B 43, 247 (1991) 13 R. K. Harris, E. D. Becker, S. M. C. De Menezes, P. 3 M. Bankay, M. Mali, J. Roos, and D. Brinkmann, Phys. Granger,R.E.Hoffman,K.W.Zilm,MagneticResonance Rev.B 50, 6416 (1994) in Chemistry 46, 582, (2008) 4 D. C. Johnston, Phys.Rev.Lett. 62, 957 (1989) 14 B. J. Suhet al., Phys.Rev. Lett.76, 1928 (1996) 5 R. E. Walstedt, B. S. Shastry, and S.-W. Cheong, Phys. 15 Y. Itoh et al., J. Phys.Soc. Jpn.67, 312 (1998) Rev.Lett. 72, 3610 (1994) 16 A. A.Gippius et al., Phys. Rev.B 59, 654 (1999) 6 J. Haase, C. P. Slichter, and G. V. M. Williams, J. Phys.: 17 S. Ohsugiet al., J. Phys. Soc. Jpn. 63, 700 (1994) Condens. Matter 21, 455702 (2009) 18 N.Curro, B.-L. Young,J. Schmalian, and D.Pines, Phys. 7 X. Zhao et al., Adv.Mater. 18, 3243 (2006) Rev.B 70, 235117 (2004) 8 N. Barisic et al., Phys.Rev. B 78, 054518 (2008) 19 D. Rybicki and et al, J. Supercond. Nov. Magn. 22, 179 9 Acomprehensiveaccountofallthedataisbeingcompleted (2009) for publication (D.Rybickiet al.) 20 V. Barzykin and D. Pines, Adv.Phys. 58, 1 (2009) 10 J. Haase and et al., Rev.Sci. Instrum.80, 073905 (2009) 21 Y. Liet al., Nature455, 372 (2008) 11 T. Meissner et al., Phys. Rev.B 83, 220517 (R) (2011) 22 Y. Liet al., Nature468, 283 (2010) 23 K. A. Mu¨ller, T. Penney, M. W. Shafer, and W. J. Fitz- patrick, Phys.Rev.Lett. 47, 138 (1981)