magnetization is parallel to the sample surface and perpendicular to the plane of incidence of light, as sketched in Fig. 3.5(c). In this thesis, the P-MOKE and L-MOKE geometries were utilized in order to determine the magnetic quantities such as HC, HEB, SQR, etc. of the probed stacks.
3.5 Chemical analysis
the energy levels is a fingerprint of each element, AES is an element specific method. As an example, considering the XYZ transition where X, Y, and Z denote the involved shells, the kinetic energies of the Auger electrons can be estimated by the formula
EXYZ=EX−EY−EZ−U(YZ), (3.7) with EX, EY, and EZ as the energy of the X, Y, and Z electron, respectively.
Furthermore, a correction term has to be added since the kinetic energy is influenced by the Coulomb interactionU(YZ)of the generated holes. In order to obtain the AES depth profile of a multilayer stack, Ar ions are employed to etch the atomic layers sequentially. Further details regarding the technique can be found in Ref. [106].
Chapter 4
Experimental
This chapter addresses results on the correlation of magnetic proper-ties with the TMR ratio in Ta-capped p-MTJs with EB as well as on the influence of different capping layers on a number of magnetic quantities and the TMR ratio. In the first section, we demonstrate the establishment of the large PEB in MnIr-based stacks. We provide a detailed analysis of the magnetic properties of HEB, HK, HC, MS, and tDL, pointing towards their tunable character by modifying the thickness of a CoFe/Ta thin layer between the MnIr and CoFeB films.
In the second section, we discuss the magnetic analysis of several trilayer CoFeB-based systems, capped with different materials (i.e. Ta, Hf, Zr, Mo). We extract the optimum parameters in terms oftCoFeB and Tannfor the establishment of PMA in the individual films. Additionally, we make a thorough analysis of the several diffusion mechanisms which may occur in the investigated systems and determine the emer-gence of PMA.
The third section focuses on the dependence of several magnetic (i.e. HEB, MsteffFM,J) and magnetotransport properties (i.e. TMR ratio) on the F thickness of the soft electrode,Tann, post annealing time, and tMgO in Ta-capped p-MTJs with EB. In addition, the correlation of the TMR ratio with the magnetic properties is discussed.
The fourth section exhibits the influence of the magnetic properties and the TMR ratio from the introduction of several capping layers
with different degree of B absorption in the vicinity of the soft CoFeB electrode. In the fifth section we investigate the role of diffusion effects in two p-MTJs with EB capped with Ta and Hf, via performing Auger measurements. In the last section we probe the efficiency of the voltage-controlled magnetic anisotropy (VCMA) effect in Ta- and Hf-capped p-MTJs with EB via examining the electric field dependence ofHC. Most of the obtained results have been published in Refs. [107, 108].
4.1 Pinned electrode stacks based on MnIr / CoFe bilayers
- 4 - 2 0 2 4
- 1
01
- 4 - 2 0 2 4
- 6 - 4 - 2 0 2 4 6
- 1
01 ( d )
( b )
( c )
Normalized Signal H ⊥ ( k O e )
N o C o F e / T a w i t h C o F e / T a
( a )
T a ( 0 . 5 0 n m ) T a ( 0 . 4 5 n m ) T a ( 0 . 4 0 n m )
ΦKerr(a.u) H ⊥ ( k O e )
T a ( 0 . 3 0 n m )
H ( k O e )
⊥
T a ( 0 . 4 0 ) n m H K
0 . 3 0 . 4 0 . 5
5678
T a ( n m )
Hk(kOe) 048
1 2
0 . 3 0 . 4 0 . 5
12HC(102 Oe) HEB(102 Oe)
Figure 4.1.(a) OOP hysteresis loops of the samples Ta/Pd/MnIr/CoFe/Ta/CoFeB/MgO (blue) and Ta/Pd/MnIr/CoFeB/MgO (red). (b) OOP hysteresis loops for variable tintTa. (c) Hysteresis loops in the OOP (blue) and IP (red) directions for the sample withtTaint=0.40 nm. (d)HK(left-axis) andHEB (right-axis) as a fuction oftTaint. The inset shows the dependence ofHContintTa.
4.1 Pinned electrode stacks based on MnIr/CoFe bilayers
Magnetic properties of MnIr-based stacks In Sec. 2.5, the necessity of the EB establishment in MTJs was analytically discussed by virtue of the enhancement of the retention time. From the magnetic standpoint, the fabrication of p-MTJs with EB involves two primary requirements. Firstly, the pinned electrode of the junction must display large EB along with lowHc, establishing a well defined plateau between the magnetic switching of the soft and pinned electrodes. In this way, the simultaneous switching of both electrodes can be prevented. Secondly, the soft and pinned electrode stacks should present high PMA to ensure a parallel (low resistance) or antiparallel (high resistance) relative orientation of the electrodes’ magnetization in the perpendicular direction. The bottom part of the junction is preferred for the development of the pinned electrode, since MnIr acts as an additional seed layer that promotes the (111) texture of the subsequent F layer and, therefore, enables the establishment of higher PMA as van Dijkenet al. reported[74].
In this thesis, the pinned electrode stack displaying large PEB and strong PMA is of the materials sequence Ta/Pd/MnIr/CoFe/Ta/CoFeB/MgO. In this stack, the presence of a CoFe/Ta dusting layer between MnIr and CoFeB plays a significant role in the satisfaction of the previously discussed magnetic criteria. Indicatively, in Fig. 4.1(a) two hysteresis loops in the OOP direc-tion are shown for the stack Ta/Pd/MnIr/CoFe/Ta/CoFeB/MgO (blue) and Ta/Pd/MnIr/CoFeB/MgO (red). An EB field equal to 730 Oe with a reduced PMA atH=0 Oe is visible for the stack without the CoFe/Ta dusting layer.
On the contrary, the emergence of PEB with anHEB=690 Oe can be realized sustaining a strong PMA atH=0 Oe for the series of stacks with the presence of CoFe/Ta interlayer.
Moreover, Fig. 4.1(b) depicts four representative hysteresis loops in the OOP direction fortintTa =0.30 nm (purple),tTaint=0.40 nm (blue),tintTa =0.45 nm (orange), tTaint =0.50 nm (green) where the tunable character of PEB as a function of thetintTa is visible. Additionally, a substantial change of the hysteresis loop’s shape can be extracted unveiling the significant influence of anisotropy while varying the tintTa.
Figure 4.1(c) presents two hysteresis loops for the stacks with tintTa =0.40 nm recorded in the IP (red) and the OOP (blue) direction, collected via AGM. The establishment of an OOP easy axis is visible when comparing the relatively higher saturation field and lower remanent magnetization of the IP hysteresis loop compared to the OOP one. TheHK, is defined as the hard-axis saturation
field and measured at the intersection of the IP and OOP hysteresis loops, as indicated by the dashed line. From the OOP (IP) loops the behaviour ofHEB(HK) is extracted and plotted against tintTa in Fig. 4.1 (d). An inverse relation betweenHKandHEBcan be realized, presenting a monotonic decrease (increase) ofHEB(HK) from 103 Oe to 1003 Oe (5029 Oe to 6204 Oe), with increasing thetTaintfrom 0.30 nm to 0.55 nm. Furthermore, from the OOP loops theHCis identified for each stack and presented in the inset of Fig. 4.1(d) as a function oftTaint. The stacks withtTaint=0.40 nm present a largerHC=191 Oe compared toHC=118 Oe for the stacks withtTaint=0.30 nm.
0 1 2 3 4 5 6
0
3 0 0 6 0 0
s l o p e : M s
( b )
MS teff FM(µemu/cm2 ) tC o F e + C o F e B ( n m )
T a ( 0 . 3 0 ) n m T a ( 0 . 4 0 ) n m T a ( 0 . 5 5 ) n m F i t t i n g
( a )
tD L
0 . 3 0 0 . 3 5 0 . 4 0 0 . 4 5 0 . 5 0 0 . 5 5
1 2 0 0 1 3 0 0 1 4 0 0
ti n tT a ( n m ) M s (emu/ccm)
0 . 0 0 . 4 0 . 8 1 . 2
t DL (nm)
Figure 4.2.(a) MstFMeff plotted against the total F thickness for the samples with tTaint=0.30 nm (red squares), tintTa =0.40 nm (green circles), tintTa =0.55 nm (blue triangles). The indicated lines represent the corresponding linear fits in whichtDL andMSare determined by the intercepts and the slopes for each case, respectively.
(b) TheMS(left-axis) andtDL(right-axis) as a function oftintTa.
In order to further elucidate the underlying mechanisms which contribute to the establishment of EB along with PMA, theMSandtDLhave been determined in a series of stacks with variabletTaint. Figure 4.2(a) illustrates the MsteffFMas a
4.1 Pinned electrode stacks based on MnIr/CoFe bilayers
function of the F thickness with the corresponding linear fit for the sample series withtTaint=0.30 nm (red squares),tTaint=0.40 nm (green circles), and tintTa =0.55 nm (blue triangles). The dead layer thicknesstDLand the saturation magnetizationMsare estimated by the intercept of the linear fit withMsteffFM=0 and the slopes of the curves, respectively. In addition, the determination ofteffFM was performed via subtracting thetFM withtDLas discussed in subsec. 2.6.3.
Figure 4.2(b) depicts the dependence ofMs(left-axis) andtDL(right-axis) on tintTa. TheMs shows a slight decrease for increasingtTaintand fortintTa ≥0.4 nm it remains constant displaying a value of 1210 emu/ccm. On the contrary, a monotonic increase of tDLis observed with increasingtintTa.
The observed behaviour ofMscould be explained through the existence of two competitive mechanisms in its final determination, as earlier discussed by Sinha et al. [109]in Ta/CoFeB/MgO layer systems. On the first hand, the determination of Msdepends on the amount of B located in the CoFeB electrode and, on the other hand, on the tDL. The deficiency of B would enhance the crystallization of CoFeB and, thus, the resultingMs. Whereas, the formation of tDLwould lead to the decrease of the determinedMsas visible from Fig. 4.2(b) fortTaint≥0.4 nm. The domination of the one mechanism over the other dictates the final result. Consequently, the formation of the dead layer obscures the effect of B absorption fortTaint≥0.4 nm, resulting in lower Ms values. Whereas, for tTaint <0.4 nm the Ms increases revealing that the mechanism of the dead layer formation is outweighed by the enhancement of B absorption.
Finally, the stacks with tTaint =0.30 nm and tTaint =0.40 nm are chosen to be the most suitable ones for the fabrication of the pinned part of the full p-MTJs. In the case of stacks with tintTa =0.30 nm, the considerably large HEB=1000 Oe is the characteristic which renders them promising candidate for their implementation in the pinned part. For stacks withtintTa =0.40 nm, although the exhibitedHEBequal to 690 Oe is smaller compared to the previous case, the obtainedHKequal to 5500 Oe is significantly larger compared to the previous ones (5000 Oe).