5.3 Solution

5.3.1 Solution with Charge Neutrality Constraint

Analytical Approach

When the charge current flows through the interface, the spin accumulation in the semiconductor appears. When K1=K2=1 (c.f. Table 5.1), the charge neutrality is restored. To solve the spin transport equations analytically for the structure shown in Figure 5.2, the general solution for the spin density deviation δs in both the FMS and the Si sides must be assumed according to [173190]

                   (- x - W- )            ( x + W-)
δsFMS  = a1ND  exp  -------2- +  a2ND exp   ----2-- ,
                       Ld                     Lu
(5.31a)

                (       W-)             (     W-)
δsSi = b1ND  exp  --x----2- + b2ND  exp  x-+--2- ,
                     Ld                    Lu
(5.31b)

where a1(a2) and b1(b2) are the constants. The spin density can be expressed as s = δs + (neql - n eql) (c.f. Equation 5.18a to Equation 5.18c). The electron concentration is n = ND. Based on these two, one can derive the up- and down-spin concentrations and the spin current density (Js) as well from Equation 5.22a and Equation 5.22b.

The external electric field () modifies the intrinsic spin diffusion length Li in any semiconductor [173]. Two distinct spin diffusion lengths (Lu is the up-stream, and Ld is the down-stream) characterize the spin motion, and those strongly depend on

Lu = ----------┌---1--------------,
               ││ (       )2
      -q|˜E-|-+  ∘   -q|˜E|--   + -1-
      2KBT         2KBT        L2i
(5.32a)

     ---------------1---------------
Ld =            ┌│  (-------)2-------,
        q|˜E |   │∘     q|˜E|       1
     - -------+      -------  +  -2-
       2KBT          2KBT        Li
(5.32b)

which is related to the intrinsic spin diffusion length Li by

L2  = LuLd.
  i
(5.33)

Li is related to the spin relaxation time via [173]

     ∘ -----
Li =   Dn τs.
(5.34)

In order to solve the transport equations, one can formulate the boundary conditions as elucidated below.

In order to derive the solution for the spin deviation and the spin current density, one must solve for 5 unknown parameters viz. a1, a2, b1, b2, and G. The spin injection efficiency can be formulated from two different approaches [173], one by the polarization of the spin current density (α) and the other by the polarization of the spin density (β)

     Js
α =  J-,
      n
(5.39)

β = s-.
    n
(5.40)

Therefore, at the silicon junction, α0 = J (0+)
-sJn- and β0 = s(0+)
n(0+).

The analytical expressions for α0 and β0 are cumbersome. The simplified expressions, valid for lower values of the bulk spin polarization P are [190]

              Ld
α0 = P ------------2----,
       Ld + (1 - P  )Lu
(5.41)

     (     Lu )
β0 =   1 - --- α0.
           Ld
(5.42)

The expression for constant G is

G  = VT ln (1 + β0P ).
(5.43)

Equation 5.42 indicates that α0 is always larger than β0 irrespective of the value of the applied field. If the electric field is very small, both the up- and down-spin diffusion lengths Lu and Ld tend to the intrinsic value and hence from Equation 5.41, α0 -P---
2-P2. This also means β0=0 and from Equation 5.43 G=0. At the strong field limit, the electrons move with the drift velocity and so does the spin polarization [173]. Ld is simply the distance over which the carriers move within the spin lifetime. Thus, when the electric field increases, Ld (  ˜
 qK|ET|
 B)Li2, L u (KBT˜-
q|E|) [173]. This makes the ratio (LLu-
 d) 0 as →∞, and both α0 and β0 tend to reach a saturation value fixed by the bulk spin polarization P at the ferromagnetic semiconductor. The maximum and minimum values of the mentioned parameters are listed in Table 5.2. Therefore, for small values of P, the spin injection efficiency in Si can be predicted analytically.


PIC

Figure 5.3: The analytically calculated spin densities in the channel are shown, when the applied voltage (Uc) is used as a parameter. The bulk spin polarization in the ferromagnetic semiconductor (FMS) is P=20%.



PIC

Figure 5.4: The spin current density (α0) and spin density (β0) injection efficiencies at the Si interface are shown as a function of the applied electric field . Linestheory and dotssimulation (P=10%).



PIC

Figure 5.5: The spin current density (α0) and spin density (β0) injection efficiencies at the Si interface as a function of the applied electric field (c.f. Figure 5.4) is shown (P=50%).




Figure 5.6: A plot of Mup = VT ln (-n↑
ne↑ql) and Mdown = VT ln (-n↓
ne↓ql) through the bar (P=10%, |qE˜|
KBT-=2μm-1 where is the applied electric field) is depicted, showing a discontinuity at the junction, which gives the term G (c.f. Equation 5.35a and Equation 5.35b). neql (n eql) is the up(down)-spin concentration at the thermal equilibrium.






Parameters

Minimum value

Maximum value




α0

2P-P2-

P

β0

0

P

G

0

VT (1 + P2)





Table 5.2: The spin injection parameters with their optimum values are shown.

Comparison between Analytical and Simulated Results

Figure 5.3 shows how spin piles up at the junction with the applied voltage Uc, when K1 = K2=1 (c.f. Table 5.1), and the bulk polarization P value is low (P = 20%). The figure depicts how the applied bias improves the spin injection efficiency.

Even though for low values of bulk polarization P the analytical model for the spin current density (spin density) injection efficiency α0 (β0) is quite precise (c.f. Figure 5.4), as soon as the value of P increases the analytical solution is no longer accurate enough and the error increases significantly (c.f. Figure 5.5). Therefore, it is also important to solve the drift-diffusion equation set numerically to lift the restriction to high values of P. It is noted in Figure 5.5 that the α0 and β0 values remain upper limited by P. Nevertheless, once up(down)-spin concentration is calculated from the simulations, the spin chemical potential drop G (c.f. Equation 5.35a and Equation 5.35b) can be obtained as described in Figure ??.

5.3.2 Solutions without Charge Neutrality Constraint


PIC

Figure 5.7: The spin density accumulation near the junction over a channel of 4μm is shown, when the current density (Jn) is fixed to 23.4MA/m2 with P=20%. K1 is used as a parameter (c.f. Table 5.1). λD represents the Debye length (c.f. Equation 5.30).



PIC

Figure 5.8: The variation of the electric potential through the channel is described, related to Figure 5.7.



PIC

Figure 5.9: The spin current density through the channel is shown, with the same conditions as in Figure 5.7. With the notations as described in Equation 5.16b and Equation 5.19b, the direction of the spin current is from the ferromagnetic semiconductor FMS towards Si.



PIC

Figure 5.10: The spin density accumulation through the channel is depicted, with the same conditions as in Figure 5.7. Both K1 and K2 are used as parameters (c.f. Table 5.1).



PIC

Figure 5.11: The spin density and the spin current density injection efficiencies (βD and αD respectively) are shown, taken at a Debye length away from the interface towards Si, with the same conditions as in Figure 5.7.


By varying K1, K2 (c.f. Table 5.1), one can remove the charge neutrality constraint. Charge injection to silicon or charge release always cause a non-zero current (hence the flow of charge) through the junction, even at the absence of an external electric field. This flow of charge causes spin accumulation/depletion at the junction. Therefore, using current as the external control parameter, rather than the applied voltage, is more convenient, because the current is always constant throughout the channel. On the other hand, if one uses voltage as the control parameter, one has to consider the potential profile in different parts of the conducting channel.

A constant current can be maintained in the channel by tuning the applied voltage Uc at certain values for K1, K2. At a constant current the distribution for the spin density deviation, the spin current density, and also the potential profile are now analyzed. Strong nonlinear effects due to the conductivity variation in the space-charge layer close to the interface cause deviations of the compensating voltage, which has to be properly considered.

It is noticed that the spin density deviation δs behaves differently at the interface and in the bulk of silicon, c.f. Figure 5.7. When K1>1, δs gradually piles up in the bulk FMS and drops down in the bulk silicon, compared to the charge neutrality condition. On the other hand when K1<1, δs gradually drops down in the FMS bulk and piles up in the Si bulk. This phenomenon happens due to the differences in the material conductivities (proportional to the doping concentration) and the electric field at the bulk, which eventually alters the spin diffusion lengths both in the FMS and Si part of the channel. However, the behavior at the interface is noted to be completely different. When K1>1 (K1<1), δs develops a dip (peak) at the FMS interface followed by a sharp peak (dip) at Si interface. These features can be correlated with the presence of charge depletion (accumulation) at the ferromagnetic/nonmagnetic interface, which results in the formation of a potential profile with a barrier for electrons (c.f. Figure 5.8). Indeed, these interface effects give rise to a very small alteration in the spin current density Js (particularly for K1=5) as observed in Figure 5.9. Nonetheless, the interface effects only persist up to the classical charge screening length λD (c.f. Equation 5.30), and completely vanish beyond that limit.

In order to investigate if the alteration of K2 has an additional effect on the observed results, one can look into Figure 5.10. It is observed that under the constraint K1=K2 the charge neutrality condition is restored. This is the reason why the additional interface effects on the spin density deviation δs is noted to be absent compared to Figure 5.7. When K1K2, the charge neutrality condition is violated at the interface. It is noted that when K2<1 (K2>1), δs develops a sharp dip (peak) in the FMS side and sharp peak (dip) in the silicon side for K1=1. However, Figure 5.10 also depicts how the spin signal remains unaltered in the silicon bulk if the value of K1 is kept fixed and the value of K2 is varied. This confirms that the tuning of the parameter K2 does not impact the bulk spin signal in silicon.

The estimated values of αD and βD, which are corresponding α and β at a distance λD away from the interface in silicon, are shown in Figure 5.11. It is revealted that αD remains greater than βD, and both are enhanced when spin is injected from a charge depleted ferromagnetic semiconductor compared to the charge neutral one. However, both αD and βD are always upper limited by the bulk spin polarization P.