Tweet

Posted on 29 June 2019

One More Way to increase the Recovery Softness of Fast High-Voltage Diodes

Free Bodo's Power Magazines!

 

 

 

By Chernikov A. A., Gubarev V. N., Stavtsev A. V., Surma A. M., and Vetrov I. Y., Proton, Electrotex ISC

 

Due to the development of the power converters based on the highvoltage IGBT and IGCT the demand for the high-voltage fast recovery diodes (FRD). The requirements to the high-voltage FRD of a new generation are as follows to assure:

• stable functioning at high values of di/dt direct current roll-off at reverse recovery (from hundreds to thousands of A/µs), and the current roll-off speed is controlled by inductivity of circuit,
• series characteristic of reverse recovery defined by
S* = tf/ts, or by

Series characteristic of reverse recovery

where ts and tf – period of the first and the second phase of reverse recovery, |(di/dt)s|max and |(di/dt)f|max – maximum absolute value of di/dt at the first and the second phase of reverse recovery.

Despite of SiC power devices fast development, still long time will find a use also silicon diodes, owing to the rather low cost and well fulfilled technology. In particular it concerns to super-power diodes (currents over several hundreds Amps, voltage over several thousand Volt).

The softer characteristic of reverse recovery of silicon FRD, the lower voltage spike at recovery of this device in the circuit with mainly inductive load, and the lower level of electromagnetic noise caused by fast changes (oscillations) of current and voltage during the switching of the module.

To increase the S-factor different special design and technological procedures are being used. The most widespread procedures are local decrease of carrier life time (τ) in the layers of semiconductor element near pn-junction [1-3] and decrease in injection efficiency of pn-junction [3-7]. At that the control of injection efficiency is achieved by increase in recombination whether in emitted layer, or on the surface of the semiconductor element adjacent to the emitter area.

Thus the variation of recombination in various areas of diode element is one of the most effective ways to influence the S-factor.

Physical preconditions and calculations results

To evaluate the influence of the recombination speed change in the layers of the diode element located at different depth, the following method was used. The process of reverse recovery of the high-voltage power diode p+-p-n--n+ element with “basic” axial distribution τ was simulated with the help of computer. A very thin test layer (r-layer) is fed into the diode element, which has a short life time compared to the “background” layer τ, which though had no big influence on concentration distribution of nonequilibrium carriers, which is typical to the “basic” element. If this test layer is fed into at various depth, and thus go through the whole depth of diode, calculated the S-factor change in each case (δS), the effect evaluation of recombination change at different depth can be achieved for that particular characteristic. The above mentioned arguments are surely true to insignificant variations of life time profile.

The results of such calculations for the element of typical power diode are presented at Figure 1. It’s clear that S increases at local decrease of τ in areas of pn-junction, and as far as the depth of r-layer close to the middle of the element, the effect gets negative what complies with the general notion.

The influence of local recombination change in different depth of diode element on S-factor.

However there is the second area near n+ layer where local decrease τ can give the positive effect again. This paradoxical at first sight result can be explained in the following way.

It is well-known that [8] at reverse recovery of p-i-n diode in i- layer of the semiconductor element two domains with high electric-field strength appear: one is adjacent to the p-layer, another to the nlayer. This happens at recovery of p+-p-n--n+ diode as well, and the strong electric field in the domain adjacent to the p+-n- junction is caused by presence of uncompensated electric charge of atoms of donor impurity and holes flowing through p+-n- junction, and in the domain adjacent to the n-- n+ boarder – by uncompensated charge of excess electrons appearing with flow of electron current through the domain. And to build the domain adjacent to the n-- n+ boarder rapid decrease in concentration of electron-hole pairs required in this area, i.e. local decrease τ near the n-- n+ boarder contributes to the building of domain with high electric-field strength.

Influence of domain adjacent to the n-- n+ boarder on the character of reverse recovery was well studied [3, 9]. It’s proven that during the process of reverse recovery the width of both domains increases which leads to domains linkage in the n- layer. As a result, “snappy recovery” occurs, i.e. the negative influence of domain building on Sfactor, adjacent to the n-- n+ boarder.

However our calculations show that in case if domain linkage doesn´t happen, the peak reverse recovery current (IrrM) and maximum rate of rise of on-state current (max |(di/dt)f|) can be decreased as a result of reassignment of applied to the diode voltage between domains.

The calculations were made with the help of "ISTOK" software used for computer modeling of semiconductor devices characteristics:
. The thickness of the silicon wafer . 640 µm, element surface . 14 cm2,
. Electron concentration in n- layer - 1.5*1013 cm-3, concentration profile of dopant impurity in p and n+ layers is shown in Figure 2a.
. Carriers´ life time close to n-- n+ border was locally decreased in comparison with its value in other parts of n- layer (Figure 2a., curve "carrier life time (1)").

Concentration profile of dopant impurity for diode element used in calculations: a. – without n’ layer; b. - with n’ layer.

Concentration profile of dopant impurity for diode element used in calculations: a. – without n’ layer; b. - with n’ layer.

The diode element was added to the loop shown in Figure 3, R=0.05 Ω, L=0.6 μH, VF=50 V, VR=-1000 V. Direct current was about 1000 A, and (di/dt)s about -1670A/μs.

The test circuit used in calculations

Characteristic curve of current and voltage at reverse recovery is shown in Figure 4, S= 5.2. In Figure 5 and Figure 6 intensity of electric field and excess-carrier density distribution through the thickness of the semiconductor element in different time periods are shown.

Characteristic curve of current (i) and voltage (U) at reverse recovery.

Distribution of electric field in diode element at different time periods.

Excess-carrier density distribution in diode element at different time periods.

According to Figure 5 it’s clear that the strong electric field area near the n-- n+ border initiates earlier than the same area near p-n junction. And exactly the strong electric field limits the peak reverse recovery current and the rate of current change in the beginning of the second phase of reverse recovery. Further smoothly decreases reverse voltage in the area of strong electric field near n-- n+ border (Un), and smoothly increases reverse voltage in the area of strong electric field near p-n junction (Up), as shown in Figure 7.

Time dependences of partial voltages in diode element at reverse recovery

Importance of the above described processes to ensure soft recovery are illustrated by the results of the comparative analysis of diode reverse recovery with similar semiconductor element, but with even distribution of τ in n- layer.

The value of τ was higher than minimal and lower than maximum τ in case with uneven distribution, and was selected to ensure he closest value of peak reverse recovery current.

In Figure 8 – Figure 10 characteristic curves of current and voltage are shown, as well as the distribution of electric field intensity and excess-carrier density according to the thickness of silicon element. Same curves for Un and Up are shown in Figure 11.

Characteristic curve of current (i) and voltage (U) at reverse recovery.

Distribution of electric field in diode element at different time periods.

Excess-carrier density distribution in diode element at different time periods.

Time dependences of partial voltages in diode element at reverse recovery

According to the above shown figures the area of strong electric field near n-- n+ border appears in this case as well, however reverse voltage in this area is relatively low. Diode reverse recovery has lower softness in this case (S=0,76).

It’s well-known that during the proton ray treatment recombination centers appear in silicon which decreases τ, and also doping centers of donor type induced by H-atoms and similar atoms of “traditional” dopants [9]. Herewith axial concentration distribution of recombination centers as well as H-induced donors has maximum near the end of proton path.

In [2, 11] it’s shown that using high-power proton irradiation through the additional screen, which is used to control the depth of proton path in the silicon element, a wide area with implanted H-atoms can be achieved. This area has a low specific resistance as a result of Hinduced donors presence. Moreover, this method allows dimensionally separate the maximums of recombination centers and H-induced donors. Thus, at Figure 12 the axial value distribution is shown, which characterizes the effective concentration of recombination defects and the concentration distribution of the implanted hydrogen at proton ray treatment of the silicon element with initial energy of 24 MeV. The concentration of the additional donors at maximum distribution can be controlled with high accuracy (error less than 1%) even at absolute values equal to the concentration of phosphorus in the nlayer (N*1013 cm-3).

The axial distribution of values (1/τ-1τ0)/(1/τ-1/τ0)max and the concentration distribution of the implanted hydrogen

Thus, such way of the proton irradiation of the silicon element from the side of the n+ layer and “hidden” super-soft n’-buffer can be built within n- layer.

The calculations show that the influence of such hidden layer on S-factor is quite ambiguous and depending on the characteristics of the semiconductor element and recovery process condition, presence of this layer can lead to decrease or to increase of S as well.

In Figure 13 and Figure. 14 characteristic curves of current and voltage are shown, as well as the electric field distribution through the thickness of the semiconductor element of diode similar to Figure 2, but containing buffer n’ layer. Concentration distribution of dopants in this layer was close to shown in Figure 12, and the maximum concentration equaled 1*1014 cm-3 at depth of 50 μm from anode surface. Corresponding curves of Un and Up are shown in Figure 15.

Characteristic curve of current (i) and voltage (U) at reverse recovery.

Distribution of electric field in diode element at different time periods.

Time dependences of partial voltages in diode element at reverse recovery

It’s clear that presence of n’ layer leads to increase of area width of strong electric field near n-- n+ boarder. On the one hand, this increases danger of area linkage of strong electric field and snappy recovery. To avoid it the thickness of the element with n’ layer should be higher than the thickness of the similar element without such layer.

The necessary condition for area of strong electric field in n- or n’ layer to appear is breaking of electrical neutrality with current flow density above some critical rates

The necessary condition for area of strong electric field in n- or n’ layer to appear is breaking of electrical neutrality with current flow density above some critical rates

q – electron charge, n0 - equilibrium density of electrons, vns – max electron speed in strong electric field. As a result within n’ layer area of strong electric field appears with relatively higher current density than for n- layer. This leads to increase of maximum surge current value at reverse recovery (IrrM) of element with n’ layer in comparison with similar element without this layer. Same thing happens when current drops in second phase of reverse recovery electric field intensity within n’ layer decreases rapidly (Figure 14) what leads to rapid decrease of Un (Figure 15).

The following factors differently influence on S-factor: increased thickness of strong electric field area helps increase of Un and S, rise delay and further rapid drop of electric field intensity within n’ layer – decreases S. For a certain example inletting into the element of n’ layer leads to decrease of S to 2.2. However with increase of |(di/dt)s| and consequently of IrrM, and VR as well (Figure 3), the situation can change.

In Figure 16 changes of S-factor are shown in dependence of VR, with reverse recovery on induction load, |(di/dt)s| = 5000 A/µs for elements with n’ layer and without it. Also the changes of (Un)max/VR for both cases are shown there. (Un)max corresponds to absolute maximum Un during the whole reverse recovery process.

Changes of S-factor and (Un)max/VR in dependence of VR

It’s clear that S is high till (Un)max/VR is about 1. Increasing VR, to some point this value is getting low, and as a result S is decreasing as well. It ought to be noted that VR when (Un)max/VR starts to decrease is higher for element with n’ layer than for similar element without it. Thus, S is higher when VR has increased value (on condition that there is no linkage of strong electric field areas).

Experimental results

Experimental diodes were produced on the base of the original neutron transmutated silicon with specific resistance of 350 Ω*cm and depth of 640 µm. The p – layer was formed by co-diffusion of B and Al at a depth of ~30 and 90 µm accordingly, and the n+ layer by the diffusion of phosphorus at a depth of 20 µm. The diameter of the silicon element of the ready-made diodes equaled 56 mm.

The diode elements were proton irradiated according to the above mentioned method for formation in the n- layer, not far from the cathode of the hidden super-soft layer, and also the layer with decreased τ adjacent to the boarder of the n+ layer. After the ray treatment the elements were annealed to activate the H-induced donors and aligning of τ. The URbr of the diodes after all treatment was 4600-4800 V.

Typical oscillograms of reverse recovery current of the experimental diodes are displayed at Figure 17. At Figure 17a. the process of recovery in the edge of low reverse voltage (100V) and current rolloff speed about 150 A/µs is shown. The value of S-factor in this case is about 1. It should be mentioned as a characteristic feature the increased period with smooth change of current near maximum.

At Figure 17b. the process of reverse recovery at VR = 1000 V is shown. The initial direct current is about 1000 A, roll-off speed is about 1600 A/µs. Typical values of S-factor were 2, no snappy-effect registered.

Typical oscillograms of experimental diode reverse recovery: a. – VR = 100 V, |(di/dt)s| = 150 A/ìs

Typical oscillograms of experimental diode reverse recovery: b. - VR = 1000 V, |(di/dt)s| = 1600 A/μs.

It has to be mentioned that current and voltage characteristic curves, which were experimentally achieved, are very close to calculation data (Figure 13).

Conclusion

The possibility in principle to develop fast p-n-n+ diode with soft recovery by local decrease of τ near the border of n+ layer was displayed.

Soft reverse recovery with high rates of anode current roll-off and reverse voltage can be assured with the help of forming the hidden buffer n’ layer in semiconductor element. The layer with decreased τ and hidden buffer n’ layer can be formed simultaneously by proton irradiation.

 

References:

1) J. Lutz and U. Scheuermann, "Advantages of the new controlled axial life-time diode," Proc. PCIM’94, Nurnberg, 1994.
2) V. A. Potaptchouk et al., "Distinctions of Lifetime Damage in Silicon Diode Layers at Various Radiation Processing: Influence on Power Losses and Softness of Reverse Recovery Characteristic," PCIM 2002, Nurnberg, 2002, pp. 293-299.
3) M. T. Rahimo and N. Y. A. Shammas, "Optimisation of the reverse recovery behaviour of fast power diodes using injection efficiency techniques and lifetime control techniques," EPE 1997, Trondheim, 1997, pp 2.99-2.104.

 

 

VN:F [1.9.17_1161]
Rating: 6.0/6 (1 vote cast)
One More Way to increase the Recovery Softness of Fast High-Voltage Diodes, 6.0 out of 6 based on 1 rating

This post was written by:

- who has written 791 posts on PowerGuru - Power Electronics Information Portal.


Contact the author

Leave a Response

You must be logged in to post a comment.