Introduction

We have already previously discovered that isoelectronic In-doping can effec-tively suppress the formation of point defects such as E1 and E2 in the DLTS measurements [75], and we also found out the new phenomenon that there is a long-term photocapacitance behavior in undoped GaN [80]. We then were cu-rious about the isoelectronic In isodoping effects on the new phenomenon. In order to understand whether or not the phenomenon still exists after the isoelec-tronic In-doping, and its influences on the logarithmic time constants, the capture cross sections, the estimated total trap concentrations (in the unit of background concentrations), and above all, the overall crystal quality, we performed the photo-capacitance measurements for both the undoped and isoelectronic In-doped GaN films.

8.2 Experiments

The undoped and In-doped GaN films were grown on (0001) sapphire sub-strates at a temperature of 1100^◦C by a low-pressure horizontal metalorganic va-por phase epitaxy (MOVPE) reactor. For undoped GaN growth, ammonia (NH3) and trimethylgallium (TMGa) were used as the N and Ga precursors with flow rates of 3 standard liter per minute and 133 µmol/min, respectively. The growth condition for iso-doped sample was almost the same as undoped one, except that a flow rate of 25.5 µmol/min of TMIn was introduced into the reactor during the sample preparation. Since the indium atom appears very difficult to enter into the solid at high growth temperatures, the resulted In/Ga ratio in the solid is less than 0.2%, as determined by secondary ion mass spectrometry. Photolu-minescence (PL) and Raman measurements further confirm isoelectronic doping properties of our GaN:In film because neither PL emission wavelength shift, nor phonon vibration mode broadening is observed. Hall measurement reveals that the electron carrier concentrations and mobilities are 4.70×10¹⁶ cm⁻³ cm⁻³, 77 cm²/V·s for undoped GaN and 8.21×10¹⁶ cm⁻³ and 88 cm²/V·s for doped one.

It seems that the Hall properties do not change significantly at this isoelectronic doping level.

Prior to the diode fabrication, the as-grown GaN sample was cleaned in suc-cessive rinses of acetone, isopropyl alcohol and D. I. water for 5 min each with ultrasonic agitation and etched by HCl: H2O= 1:1 for 10 min. Ni and Al

met-als were then deposited on the front of GaN films to form Schottky and ohmic contacts by electron-gun evaporation through metal masks.

For the measurements of photocapacitance, the Schottky diode was illumi-nated by a nominal 60 W tungsten lamp for 5 min at zero bias voltage. The sample was then used for the measurement of transient behavior at room temper-ature using a HP4194 impedance analyzer at different dark waiting times from 5 min to 12 h. For each measurement, a reverse bias of -1 V square voltage with duration of ∼1.5 s was applied for the purpose of emptying the carrier in the deple-tion region. Pulses with a test frequency of 10 kHz and 100 meV oscilladeple-tion level were employed to record the small signal diode capacitance. The trigger duration used was sufficiently long to attain the respective steady state capacitance value, CS(t), so that long-term photocapacitance behavior could be obtained. Attention was paid to ensure that the time interval between the consecutive measurements was at least 4 min. Under these circumstances, any possible intervention between the measurements can be minimized to a negligible extent.

8.3 Results and Discussion

Figures 8.1(a) to 8.1(d) shows the transient capacitance of Ni Schottky diodes on both undoped and In-doped GaN epilayers measured at room temperature probed at different time, t, after the illumination was switched off. For each measurement in Fig. 8.1 when a square triggering voltage is applied, the

capaci-tance signal increases exponentially and saturates to a steady-state value within a duration of ∼0.2 s. The uprising portion has been attributed primarily to the point defects associated with the deep level of ∼0.6 eV below the conduction, as previously revealed by deep level transient spectroscopy measurements.[75] This particular trap cannot be due to its fast dynamic response in nature, cannot be considered responsible for the long-term photocapacitance behavior. It is worth noting that the steady-state capacitances CS(t) were recorded respectively in 5 min and 1 h units and also plotted in shifted ordinates for both undoped and In-doped GaN Schottky diodes. The heights of CS(t) on both sides of the figures are virtually the same for all different time scales used.

Figure 8.2 shows the plot of the time dependence of the steady-state values of the photocapacitance, CS(t), of Ni Schottky diodes on both undoped and In-doped GaN epilayers. We observed that CS(t) decreases logarithmically as a function of t with its value dropping from 116 pF at t = 1 min to 106 pF at t = 36 h for undoped GaN, and from 200 pF at t = 1 min to 188 pF at t = 66 h for In-doped GaN. The vertical offset in between was due to the slight difference of concentrations of net ionized impurity in the space charge region. As can be seen in Fig. 8.2, CS(t)is found to maintain its logarithmic decay behavior as a function of t when small quantities of In are doped into the epilayer.

From eq. (7.10) and the CS(t) versus t curve in Fig. 8.2, we then obtained nT(t) of each t and plotted in Fig. 8.3 and Fig. 8.4. Figure 8.3 shows the ratio

0.0 0.3 0.6 0.9 1.2

Figure 8.1: The transient capacitance response of the undoped and In-doped GaN during the applications of square voltage recorded at different waiting time, t, at room temperature.

1 10 100 1000 0

50 100 150 200 250 300

In-doped Undoped

C S(t) (pF)

Time (min)

Figure 8.2: The steady state value of transient capacitance, CS(t) versus waiting time t at room temperature for undoped and In-doped GaN.

-500 0 500 1000 1500 2000 2500 3000 3500 4000 4500 -0.02

0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20

Time (min) n T(t)/ N D

Undoped GaN In doped GaN

Figure 8.3: The ratio of nT(t)and ND versus waiting time t at room temperature for undoped and In-doped GaN films in linear scale.

1 10 100 1000 0.00

0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20

Time (min) n T(t)/ N D

Undoped GaN In doped GaN Simulation

Figure 8.4: The ratio of nT(t)and ND versus waiting time t at room temperature for undoped and In-doped GaN films in log scale.

of nT(t) and ND versus waiting time t at room temperature for undoped and In-doped GaN films in linear scale. As can be seen in the figure, the NT/ND

(obtained by the maximum value of nT/ND as t goes to the infinity, that is, the time larger than 2500 min) were improved from 18% to 13%. The isoelectronic In doping indeed improved the crystal quality through the suppression on the formation of dislocation-related trap.

Figure 8.4 shows the ratio of nT(t) and ND versus waiting time t at room temperature for undoped and In-doped GaN films in log scale. According to the simulation method we previously revealed,[80] from ratio of the retrieved slope and intercept of nT(t) = a + b ln t, using eq. (7.8) and eq. (7.9), we obtained that the time constants and electron capture cross sections are 34.5 s and 2.89×10⁻²⁷ cm² for undoped GaN and 13.7 s and 3.92×10⁻²⁷ cm² for In-doped ones.

The enlarged electron capture cross section and the shortened logarithmic time constant after the incorporation of In atoms are consistent with the reduction of the concentration of related traps, since the suppression of dislocation-related traps would introduce a smaller repulsive Coulomb potential, which would make the traps capture the electrons more easily and form the cylindrical poten-tial around the dislocations more easily, and then introduce the larger electron capture cross section and smaller logarithmic time constant after the isoelectronic In doping.

8.4 Summary

In summary, we have measured the dynamic response of photocapacitance of Schottky diodes made of both undoped and isoelectronically In-doped GaN samples at various waiting time t after illumination was switched off. Experi-mental results reveal that when small quantities of In atoms are added into the epilayer, the logarithmic time constant was improved from 34.5 to 13.7 s, and the formation of the concentration of the affiliated dislocation traps can be effectively suppressed. This study confirms that isoelectronic doping technique is a simple, yet useful method in reducing the concentration of electron-hole recombination centers at dislocations. Application of this technique may bring an improvement in GaN device performance, particularly for those films grown with large defect concentration.

CHAPTER 9

SURFACE TREATMENT EFFECTS ON P -GAN SCHOTTKY DIODES

9.1 Introduction

The wide-bandgap gallium nitride is a promising material for high-temperature, high-power GaN-based field-effect devices owing to its peculiar properties, such as high electron velocity, good thermal conductivity, high breakdown field, high sheet charge density, large conduction band discontinuity and piezoelectric effect.[81, 82]

Because of the use of field effect nature in these devices, the Schottky performance is without question one of the most important issues in implementation of these devices. Up to date, most of relevant works reported has focused on the Schottky properties of n-type GaN film, very few literatures have been addressed on that of p-type GaN film. For the case of p-GaN Schottky diode, T. Mori et al. have car-ried out the study as early as 1996 and reported a barrier height of ∼0.65 eV.[83]

This result is considered to be much lower than the expected value calculated from Schottky approximation, indicating that non-ideal current mechanisms are predominant in device carrier transport. Since the rectifying characteristics rely greatly on the surface properties of the semiconductor, several surface

pretreat-ment methods including the use of HF, HCl solvents, or complex combination of HCl, HF and (NH4)2S chemical cleaning steps, have been employed to remove the native oxide and modify the surface electrical properties.[83, 84, 85] Nevertheless, only limited successes have been achieved. Most of p-GaN Schottky diodes still exhibits very leaky characteristics unless it is made on a low Mg-doped GaN epi-film, which is in turn taken at the expense of high device series resistance and low operation current. In this study, we have made p-type Schottky diodes on GaN film using different surface cleaning process. Our experiment shows that when the substrate is subjected to ammonia sulfide surface pretreatment process good p-GaN Schottky contacts can be obtained on a nominal p-GaN epitaxial layer having a hole concentration of 2.5×10¹⁷ cm⁻³. Calculation also indicates that the sulfide-treated p-GaN sample is nearly free from any surface contaminations. The resulted surface state density is as low as 2.004×10¹² cm⁻²eV⁻¹.

9.2 Experiments

The epitaxial Mg-doped GaN film used in this experiment was 1-µm thick, grown by low-pressure horizontal metalorganic vapor phase epitaxy (MOVPE).

The corresponding Hall concentration and mobility were measured to be 2.5×10¹⁷ cm⁻³ and 8 cm²/V·s, respectively, after the thermal annealing at 750 ^oC for 40 min. Mirror-like surface morphology is observed for this film and the associated root-mean-square roughness is about 8 Å as revealed by atomic force microscopy

measurement. Prior to the diode fabrication, the wafer was cut into several pieces of samples of approximate area of 1 cm². All of the samples were degreased in successive rinses of acetone, isopropyl alcohol and de-ionized water for 5 min each with ultrasonic agitation, and dried with nitrogen gas. Several different surface treatments were then employed for these samples. They were immersed either in an HCl solution at room temperature for 4 min or in a boiled (NH4)2S solution for 10 min. These samples were subsequently rinsed in de-ionized water, blown dry by nitrogen gas again, and quickly loaded into an electron beam evaporator.

Ni metal was then evaporated onto the sample surface through the metal mask to form the Schottky contacts with a diameter of ∼100 µm, and to form the stripe ohmic contacts at the same time due to the use of large surface area in this study. To verify the surface treatment effects, we also made Schottky diodes from an Mg-doped GaN film without any inorganic surface cleaning process here for comparison.

9.3 Results and Discussion

The I-V characteristics of these samples were measured with a HP4145B semi-conductor analyzer and the C-V profiles were measured by a HP4194A impedance analyzer at a frequency of 300Hz. The effective barrier height φ_B and ideality fac-tor n of the Schottky diode can be obtained from the I-V characteristics, assuming carrier transport is dominated by thermionic emission. The I-V characteristics in

the forward bias is given by[86]

where J is the current density, Js the saturation current density, the Richard-son constant ( 96 Acm⁻²K⁻² for p-GaN),[85] T the diode temperature, Rs series resistance, q the electron charge, n ideality factor and k the Boltzmann’s constant.

In Figs. 9.1(a) and 9.1(b), we show the typical room-temperature current density-voltage characteristics of the above Ni Schottky diodes in both linear and log scales. As can be seen in Fig. 9.1(a), almost no rectifying I-V characteristics can be observed for the untreated p-GaN Schottky diode. Although HCl surface treatment gives better results in Schottky properties, considerably high leakage current is still observed at reverse bias, very similar to case of the p-GaN Schot-tky study reported by K. N. Lee et al.[87] For these two samples, the slopes of reverse leakage current vs. bias voltage in log-scale are found to equal approxi-mately 2, showing that the leakage currents are proportional to the device surface area. Such a fact implies that the reverse carrier transport in these two types of samples is influenced significantly by the surface states[86] between the metal and semiconductor interface, which conceivably is one of the major causes responsible for the degradation of such types of p-GaN Schottky diodes.

On the other hand, well-behaved rectifying characteristics were obtained in

-1.0 -0.5 0.0 0.5 1.0

Figure 9.1: Room temperature current density-voltage characteristics of Ni-GaN Schottky diodes with various surface treatments in both linear and log scales.

(NH4)2S-treated p-GaN Schottky diode. As compared to the previous diodes, we can find there are dramatic effects on the diode properties as sulfide treat-ment is used in preparing p-GaN Schottky diode. The calculated ideality factor n, saturation current JS and barrier height φ_b are 1.14, 1.53×10⁻⁹ A/cm² and 0.94 eV, respectively, much better than those prepared by the previous methods.

Moreover, the resulted reverse leakage current is rather low, down to a value of

∼5×10⁻⁸A/cm², which represents one of the best values ever reported for p-type GaN Schottky diode. Different from the results by X. A. Cao et al.[85], our exper-iment demonstrates that the sulfide surface cleaning process is a viable method for preparing good quality p-type Schottky diode.

From the above discussions, particularly the facts of proportionality between the leakage current and diode surface area, together with very leaky current be-havior in untreated and HCl-treated diodes, we can infer that the surface of p-GaN semiconductor is by nature unstable itself. That is without proper surface treat-ment, the interface of metal-semiconductor could usually be covered by a large concentration of surface defects due to the dangling bonds on the p-GaN surface.

In order to investigate the surface states more qualitatively, we thus conducted a C-V measurement for these samples. Figure 9.2 shows the C-V characteristics of Ni Schottky diodes with different surface treatments measured at room tem-perature. As can be seen in the figure, the linearity of 1/C² as a function of reverse bias voltage is held for the sulfide-treated and the HCl-treated samples

-1.0 -0.8 -0.6 -0.4 -0.2 0.0 -2.0x10²¹

0.0 2.0x10²¹ 4.0x10²¹ 6.0x10²¹ 8.0x10²¹ 1.0x10²² 1.2x10²² 1.4x10²² 1.6x10²²

1/C2 (F -2 )

Voltage(V)

Untreated HCl-treated Sulfide-treated

Figure 9.2: Room temperature capacitance-voltage characteristics of Ni-GaN Schottky diodes with various surface treatments.

only, but is invalid for untreated one. The capacitance against applied voltage for the Schottky diode can be expressed as

1

C² = 2(Vbi+ VR− kT/q)

²sqNaA² , (9.3)

where Vbiis the built-in potential, VRthe reverse bias voltage, ²s= 9.5 ²0for GaN, Nathe acceptor concentration, A is the area of Schottky diode. From the intercept on the voltage axis, the barrier heights can be also determined:

φ_Bp= Vbi+ (Ef − E^v)/q− ∆φm = χ + Eg/q− φm− ∆φBp (9.4)

Here Ef and EV are the Fermi level and valence band, and are the electron affinity (4.1 eV) and bandgap (3.4 eV) for GaN, φ_m is the Ni metal work function (5.2 eV) and ∆φ_Bp represents the barrier lowering due to the image force, which is given by

Using these equations, we can have the acceptor concentrations and barrier heights of 1.08×10¹⁸, 1.56×10¹⁸ cm⁻³ and 0.970, 2.190 eV for HCl-treated and (NH4)2 S-treated samples, respectively. One can note that as the C-V barrier height of sulfide-treated sample is twice the value of that of HCl-treated one, confirming the significance of sulfide treatment in preparing p-GaN Schottky diode. After adding up the measured Schottky barrier height φ_Bp with ∆φ_Bp (0.093 eV at reverse bias of one volt), we have a value of 2.283 eV, very close to that of ideal barrier height 2.3 eV derived from Schottky approximation.

The highly coincidence between the experimental and theoretical values strongly suggests that the surface of sulfide treated p-GaN is nearly free from any conta-minations. Indeed, we have found the density of surface states is approximately of 2.004×10¹² cm⁻²·eV⁻¹ in accord with the following formula we developed fol-lowing the model by Cowley and Sze[86]

φ_Bp+ ∆φ_Bp= b2[Eg/q− (φm− χ)] + (1 − b²)(Eg/q− φ0), (9.6)

and

Ds = (b2− 1)²ⁱ

b2δq² , (9.7)

, by assuming that the sulfide cleaned p-GaN surface has a thin interfacial layer with δ = 5 Å, ²i = 3.8 ²0, and φ₀ = 2.334 eV.[86, 88, 89, 90] Here, δ is the thickness of the interfacial layer and ²i is its permittivity, and φ₀ represents the neutral energy level for the surface states. Such a value is considerably lower than those of most III-V semiconductor compounds, such as Si, GaP and GaAs, which lying in the range from 2.7×10¹³to 12.5×10¹³ cm⁻²·eV⁻¹. On the contrary, the resulted surface state density for HCl-treated sample is quite large. The corresponding value is as high as 1.020×10¹⁵ cm⁻²·eV⁻¹ (δ = 10 Å and ²i = 10.2

²0, and φ₀ = 2.334 eV),[85, 86, 91] comparable to the maximum dangling bond density 3.404×10¹⁵ cm⁻² eV⁻¹ that can be allowed for GaN surface.

We believe such a distinct surface density reduction in sulfide-treated p-GaN diode is contributed primarily to the surface passivation effects imposed by sulfur atoms. It has been reported that when the (NH4)2S solution is introduced during

the cleaning process for GaN, it will not only produce a low coverage of oxygen on the surface, but the sulfur atoms can also fill in the nitrogen vacancies and react with the gallium atoms to form stable Ga-S bonds to prevent from further oxidation in the air.[89] Consequently, a virtually contamination-free surface can be obtained for the p-type GaN film after the sulfide pretreatment surface process.

Although, the HCl process can also effectively remove the native oxide, it seems not necessary to warrant a low concentration of surface defects on p-type GaN sample. S. W. King et al.[92] have pointed out that the exposure of GaN surface to HCl solution could inevitably leave significant amounts of residual Cl contami-nants and C-H bonded carbons on the surface. That may explain partially for the extraordinary high density of surface states observed in our HCl-treated p-GaN Schottky diode.

9.4 Summary

In summary, we have investigated the surface treatment effects on the I-V and C-V characteristics of p-GaN Schottky diodes using different cleaning steps.

Poor rectifying characteristics were observed on both untreated and HCl-treated samples. When the sample was undergone (NH4)2S treatment process, signifi-cantly improved Schottky behaviors could be obtained. The resulted ideality fac-tor, reverse leakage current, I-V barrier height and C-V barrier height are 1.14, 1.53×10⁻⁹A/cm², 0.94 eV and 2.19 eV, respectively, indicating the good quality of

our p-type GaN Schottky contact. Furthermore, we have also found that (NH4)2S surface pretreatment has a profound effect in the reduction of surface state density for GaN. The corresponding surface state density can be reduced to a value as low as 2.004×10¹²cm⁻²eV⁻¹, which is about a three-order of magnitude improved

our p-type GaN Schottky contact. Furthermore, we have also found that (NH4)2S surface pretreatment has a profound effect in the reduction of surface state density for GaN. The corresponding surface state density can be reduced to a value as low as 2.004×10¹²cm⁻²eV⁻¹, which is about a three-order of magnitude improved