Current-Capacitance characteristics

The samples were measured for determining the C-V characteristics. The voltage was swept from accumulation (−13 V ) to inversion (10 V ) and back. The crossing between regions induces a band flattening of the structure and the capacitance falls to the flat band value (Cf b). The voltage that induces this state is called flat band

voltage (Vf b). The measurement of Vf band the hysteresis provide information about

the charge density in the AlN film and the injected charge during the measurement. A frequency sweep was performed for determining the optimal measurement conditions. The capacitance values measured for accumulation and inversion are different but tend to a similar value as the measurement frequency increases. Nev- ertheless, the maximum capacitance drop at the flat band condition is obtained for f = 100 KHzand thus this frequency was selected for performing the measurements.

Charge density determination

The measurement and interpretation of C-V curves is a standard method for semi- conductor characterization. Details on the measurement principles may be consulted

at [165]. In what follows a brief summary of the calculations used for the AlN characterization is provided.

The flat band capacitance is given by Cf b =

0AlNSi

AlNLD + SidAlN

(5.2) where 0, AlN and Si are the electrical permittivities of vacuum, AlN and Si,

respectively, dAlN the thickness of the AlN film and LD the Debye length. The Debye

length is calculated as LD = s Si0KBT q2_N Si (5.3) were KB is the Boltzmann constant, T the temperature in degrees Kelvin, q the

electron charge and NSi the dopant density of the silicon substrate. As all of these

are known quantities, LD is calculated to be approximately 40 µm.

Taking the calculated value for LD and the theoretical values for the electrical

permittivities, Cf b can be calculated. The variation due to the thickness of the AlN

layer is small (as dAlN  LD) and will be masked by the uncertainties in other factors,

such as the exact dopant density of Si or the permittivities. Hence a constant average value of Cf bis considered regardless the AlN thickness,

Cf b≈ 2.59 · 10−6 F/m2 (5.4)

Taking into account the contact surface of the Hg probe results in

Cf b≈ 4.4 pF (5.5)

The trapped charge in the dielectric layer is given by Vf b following

NF =

 φms− Vf b

qdAlN



AlN0 (5.6)

where NF is the density of charge trapped in the dielectric layer and φms is the

work function of the metal-semiconductor contact.

It has to be stated that all the measurement provided a minimum capacitance very close to the 4.4 pF value. The small variations could be attributed to uncertainties in the values used in the calculation, but their magnitude has to be small as the dispersion of the measured Cf b is also small (< 5%). Therefore, the value of the

value of  = 9.14.

Additionally, large capacitance hysteresis effect was observed when performing the voltage sweep. The hysteresis turns counterclockwise so the effect is attributed to hole carrier injection. When starting from the accumulation regions, holes are injected in the dielectric, and when starting from inversion electrons are injected. Therefore, the mean value of Vf bhas been taken as an estimator of the charge trapped

in the dielectric and the hysteresis amplitude as an estimator of the charge injection.

Measurement results

A large dispersion in the measured values has been observed. The values of trapped charge are high, between 8 · 1011_cm−2 _{and 4 · 10}12_cm−2_{, and comparable}

to other published results [164, 166]. The hysteresis effect produces densities of in- jected charges of about an order of magnitude lower, from 1·1011_cm−2 _{to 1·10}12_cm−2_,

but following the trends of NF.

The only clear trend that has been identified is the fall of the trapped and injected charge values as the layers get thicker. Indeed, the trapped charge value falls from the range 2 − 4 · 1012_cm−2 _{to 0.8 − 2 · 10}12_cm−2 _{as the AlN thickness evolves from}

60 nmto 120 nm. This effect suggests that the majority of the carrier is trapped at the interface [164, 166] and that some of it is removed during growth of thicker layers, probably due to defect annealing by self-heating effect.

The rest of the data suggests a small decrease of the charge density value for samples with greater texture, altough it is difficult to confirm due to the scarce data points and the large dispersion of NF. What is clear from the data is that the disper-

sion of NF for samples deposited with dts = 70 mm is lower than for dts = 50 mm.

The measured values of the samples deposited at dts = 70 mm tend to concentrate

Constant Value Hg probe diameter 735 µm Hg work function 4.53 eV χAlN 0.6 eV φms 3.93 eV AlN 9.14 Si 11.7 T 300 K NSi 1016cm−3

around 2.6 · 1012_cm−2 _{for the thinner films and around 1.2 · 10}12_cm−2 _{for the thicker.}

In summary, it can be concluded from the C-V characterization that:

• A large density of trapped charge is preset at the deposited films. It is believed to be mostly near the interface at the adaptation layer.

• The trapped charge decreases with increasing thickness probably due to defect annealing.

• The thicker adaptation layer for the samples deposited at dts = 50 mmcause a

larger dispersion in the calculated Nf b.

• The density of injected charge is large, although an order of magnitude lower than NF, and follows the trends with the deposition conditions of the trapped

charge in the layer.